Implement Wizard for Creating Classification Models (#20622)

* Implement extraction of images for classification state models

* Add object classification dataset preparation

* Add first step wizard

* Update i18n

* Add state classification image selection step

* Improve box handling

* Add object selector

* Improve object cropping implementation

* Fix state classification selection

* Finalize training and image selection step

* Cleanup

* Design optimizations

* Cleanup mobile styling

* Update no models screen

* Cleanups and fixes

* Fix bugs

* Improve model training and creation process

* Cleanup

* Dynamically add metrics for new model

* Add loading when hitting continue

* Improve image selection mechanism

* Remove unused translation keys

* Adjust wording

* Add retry button for image generation

* Make no models view more specific

* Adjust plus icon

* Adjust form label

* Start with correct type selected

* Cleanup sizing and more font colors

* Small tweaks

* Add tips and more info

* Cleanup dialog sizing

* Add cursor rule for frontend

* Cleanup

* remove underline

* Lazy loading

.cursor/rules/frontend-always-use-translation-files.mdc

@@ -0,0 +1,6 @@
+---
+globs: ["**/*.ts", "**/*.tsx"]
+alwaysApply: false
+---
+
+Never write strings in the frontend directly, always write to and reference the relevant translations file.

docs/docs/configuration/custom_classification/object_classification.md

@@ -12,7 +12,18 @@ Object classification models are lightweight and run very fast on CPU. Inference
 Training the model does briefly use a high amount of system resources for about 1–3 minutes per training run. On lower-power devices, training may take longer.
 When running the `-tensorrt` image, Nvidia GPUs will automatically be used to accelerate training.
 
-### Sub label vs Attribute
+## Classes
+
+Classes are the categories your model will learn to distinguish between. Each class represents a distinct visual category that the model will predict.
+
+For object classification:
+
+- Define classes that represent different types or attributes of the detected object
+- Examples: For `person` objects, classes might be `delivery_person`, `resident`, `stranger`
+- Include a `none` class for objects that don't fit any specific category
+- Keep classes visually distinct to improve accuracy
+
+### Classification Type
 
 - **Sub label**:
 

docs/docs/configuration/custom_classification/state_classification.md

@@ -12,6 +12,17 @@ State classification models are lightweight and run very fast on CPU. Inference
 Training the model does briefly use a high amount of system resources for about 1–3 minutes per training run. On lower-power devices, training may take longer.
 When running the `-tensorrt` image, Nvidia GPUs will automatically be used to accelerate training.
 
+## Classes
+
+Classes are the different states an area on your camera can be in. Each class represents a distinct visual state that the model will learn to recognize.
+
+For state classification:
+
+- Define classes that represent mutually exclusive states
+- Examples: `open` and `closed` for a garage door, `on` and `off` for lights
+- Use at least 2 classes (typically binary states work best)
+- Keep class names clear and descriptive
+
 ## Example use cases
 
 - **Door state**: Detect if a garage or front door is open vs closed.

frigate/api/app.py

@@ -387,7 +387,8 @@ def config_set(request: Request, body: AppConfigSetBody):
         old_config: FrigateConfig = request.app.frigate_config
         request.app.frigate_config = config
 
-        if body.update_topic and body.update_topic.startswith("config/cameras/"):
+        if body.update_topic:
+            if body.update_topic.startswith("config/cameras/"):
                 _, _, camera, field = body.update_topic.split("/")
 
                 if field == "add":
@@ -401,6 +402,13 @@ def config_set(request: Request, body: AppConfigSetBody):
                     CameraConfigUpdateTopic(CameraConfigUpdateEnum[field], camera),
                     settings,
                 )
+            else:
+                # Handle nested config updates (e.g., config/classification/custom/{name})
+                settings = config.get_nested_object(body.update_topic)
+                if settings:
+                    request.app.config_publisher.publisher.publish(
+                        body.update_topic, settings
+                    )
 
     return JSONResponse(
         content=(

frigate/api/classification.py

@@ -3,7 +3,9 @@
 import datetime
 import logging
 import os
+import random
 import shutil
+import string
 from typing import Any
 
 import cv2
@@ -17,6 +19,8 @@ from frigate.api.auth import require_role
 from frigate.api.defs.request.classification_body import (
     AudioTranscriptionBody,
     DeleteFaceImagesBody,
+    GenerateObjectExamplesBody,
+    GenerateStateExamplesBody,
     RenameFaceBody,
 )
 from frigate.api.defs.response.classification_response import (
@@ -30,6 +34,10 @@ from frigate.config.camera import DetectConfig
 from frigate.const import CLIPS_DIR, FACE_DIR
 from frigate.embeddings import EmbeddingsContext
 from frigate.models import Event
+from frigate.util.classification import (
+    collect_object_classification_examples,
+    collect_state_classification_examples,
+)
 from frigate.util.path import get_event_snapshot
 
 logger = logging.getLogger(__name__)
@@ -159,8 +167,7 @@ def train_face(request: Request, name: str, body: dict = None):
     new_name = f"{sanitized_name}-{datetime.datetime.now().timestamp()}.webp"
     new_file_folder = os.path.join(FACE_DIR, f"{sanitized_name}")
 
-    if not os.path.exists(new_file_folder):
+    os.makedirs(new_file_folder, exist_ok=True)
-        os.mkdir(new_file_folder)
 
     if training_file_name:
         shutil.move(training_file, os.path.join(new_file_folder, new_name))
@@ -701,13 +708,14 @@ def categorize_classification_image(request: Request, name: str, body: dict = No
             status_code=404,
         )
 
-    new_name = f"{category}-{datetime.datetime.now().timestamp()}.png"
+    random_id = "".join(random.choices(string.ascii_lowercase + string.digits, k=6))
+    timestamp = datetime.datetime.now().timestamp()
+    new_name = f"{category}-{timestamp}-{random_id}.png"
     new_file_folder = os.path.join(
         CLIPS_DIR, sanitize_filename(name), "dataset", category
     )
 
-    if not os.path.exists(new_file_folder):
+    os.makedirs(new_file_folder, exist_ok=True)
-        os.mkdir(new_file_folder)
 
     # use opencv because webp images can not be used to train
     img = cv2.imread(training_file)
@@ -756,3 +764,43 @@ def delete_classification_train_images(request: Request, name: str, body: dict =
         content=({"success": True, "message": "Successfully deleted faces."}),
         status_code=200,
     )
+
+
+@router.post(
+    "/classification/generate_examples/state",
+    response_model=GenericResponse,
+    dependencies=[Depends(require_role(["admin"]))],
+    summary="Generate state classification examples",
+)
+async def generate_state_examples(request: Request, body: GenerateStateExamplesBody):
+    """Generate examples for state classification."""
+    model_name = sanitize_filename(body.model_name)
+    cameras_normalized = {
+        camera_name: tuple(crop)
+        for camera_name, crop in body.cameras.items()
+        if camera_name in request.app.frigate_config.cameras
+    }
+
+    collect_state_classification_examples(model_name, cameras_normalized)
+
+    return JSONResponse(
+        content={"success": True, "message": "Example generation completed"},
+        status_code=200,
+    )
+
+
+@router.post(
+    "/classification/generate_examples/object",
+    response_model=GenericResponse,
+    dependencies=[Depends(require_role(["admin"]))],
+    summary="Generate object classification examples",
+)
+async def generate_object_examples(request: Request, body: GenerateObjectExamplesBody):
+    """Generate examples for object classification."""
+    model_name = sanitize_filename(body.model_name)
+    collect_object_classification_examples(model_name, body.label)
+
+    return JSONResponse(
+        content={"success": True, "message": "Example generation completed"},
+        status_code=200,
+    )

frigate/api/defs/request/classification_body.py

@@ -1,17 +1,31 @@
-from typing import List
+from typing import Dict, List, Tuple
 
 from pydantic import BaseModel, Field
 
 
 class RenameFaceBody(BaseModel):
-    new_name: str
+    new_name: str = Field(description="New name for the face")
 
 
 class AudioTranscriptionBody(BaseModel):
-    event_id: str
+    event_id: str = Field(description="ID of the event to transcribe audio for")
 
 
 class DeleteFaceImagesBody(BaseModel):
     ids: List[str] = Field(
         description="List of image filenames to delete from the face folder"
     )
+
+
+class GenerateStateExamplesBody(BaseModel):
+    model_name: str = Field(description="Name of the classification model")
+    cameras: Dict[str, Tuple[float, float, float, float]] = Field(
+        description="Dictionary mapping camera names to normalized crop coordinates in [x1, y1, x2, y2] format (values 0-1)"
+    )
+
+
+class GenerateObjectExamplesBody(BaseModel):
+    model_name: str = Field(description="Name of the classification model")
+    label: str = Field(
+        description="Object label to collect examples for (e.g., 'person', 'car')"
+    )

frigate/data_processing/real_time/custom_classification.py

@@ -53,9 +53,17 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
         self.tensor_output_details: dict[str, Any] | None = None
         self.labelmap: dict[int, str] = {}
         self.classifications_per_second = EventsPerSecond()
+
+        if (
+            self.metrics
+            and self.model_config.name in self.metrics.classification_speeds
+        ):
             self.inference_speed = InferenceSpeed(
                 self.metrics.classification_speeds[self.model_config.name]
             )
+        else:
+            self.inference_speed = None
+
         self.last_run = datetime.datetime.now().timestamp()
         self.__build_detector()
 
@@ -83,9 +91,11 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
 
     def __update_metrics(self, duration: float) -> None:
         self.classifications_per_second.update()
+        if self.inference_speed:
             self.inference_speed.update(duration)
 
     def process_frame(self, frame_data: dict[str, Any], frame: np.ndarray):
+        if self.metrics and self.model_config.name in self.metrics.classification_cps:
             self.metrics.classification_cps[
                 self.model_config.name
             ].value = self.classifications_per_second.eps()
@@ -223,9 +233,17 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         self.detected_objects: dict[str, float] = {}
         self.labelmap: dict[int, str] = {}
         self.classifications_per_second = EventsPerSecond()
+
+        if (
+            self.metrics
+            and self.model_config.name in self.metrics.classification_speeds
+        ):
             self.inference_speed = InferenceSpeed(
                 self.metrics.classification_speeds[self.model_config.name]
             )
+        else:
+            self.inference_speed = None
+
         self.__build_detector()
 
     @redirect_output_to_logger(logger, logging.DEBUG)
@@ -251,9 +269,11 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
 
     def __update_metrics(self, duration: float) -> None:
         self.classifications_per_second.update()
+        if self.inference_speed:
             self.inference_speed.update(duration)
 
     def process_frame(self, obj_data, frame):
+        if self.metrics and self.model_config.name in self.metrics.classification_cps:
             self.metrics.classification_cps[
                 self.model_config.name
             ].value = self.classifications_per_second.eps()

frigate/embeddings/maintainer.py

@@ -9,6 +9,7 @@ from typing import Any
 
 from peewee import DoesNotExist
 
+from frigate.comms.config_updater import ConfigSubscriber
 from frigate.comms.detections_updater import DetectionSubscriber, DetectionTypeEnum
 from frigate.comms.embeddings_updater import (
     EmbeddingsRequestEnum,
@@ -95,6 +96,9 @@ class EmbeddingMaintainer(threading.Thread):
                 CameraConfigUpdateEnum.semantic_search,
             ],
         )
+        self.classification_config_subscriber = ConfigSubscriber(
+            "config/classification/custom/"
+        )
 
         # Configure Frigate DB
         db = SqliteVecQueueDatabase(
@@ -255,6 +259,7 @@ class EmbeddingMaintainer(threading.Thread):
         """Maintain a SQLite-vec database for semantic search."""
         while not self.stop_event.is_set():
             self.config_updater.check_for_updates()
+            self._check_classification_config_updates()
             self._process_requests()
             self._process_updates()
             self._process_recordings_updates()
@@ -265,6 +270,7 @@ class EmbeddingMaintainer(threading.Thread):
             self._process_event_metadata()
 
         self.config_updater.stop()
+        self.classification_config_subscriber.stop()
         self.event_subscriber.stop()
         self.event_end_subscriber.stop()
         self.recordings_subscriber.stop()
@@ -275,6 +281,46 @@ class EmbeddingMaintainer(threading.Thread):
         self.requestor.stop()
         logger.info("Exiting embeddings maintenance...")
 
+    def _check_classification_config_updates(self) -> None:
+        """Check for classification config updates and add new processors."""
+        topic, model_config = self.classification_config_subscriber.check_for_update()
+
+        if topic and model_config:
+            model_name = topic.split("/")[-1]
+            self.config.classification.custom[model_name] = model_config
+
+            # Check if processor already exists
+            for processor in self.realtime_processors:
+                if isinstance(
+                    processor,
+                    (
+                        CustomStateClassificationProcessor,
+                        CustomObjectClassificationProcessor,
+                    ),
+                ):
+                    if processor.model_config.name == model_name:
+                        logger.debug(
+                            f"Classification processor for model {model_name} already exists, skipping"
+                        )
+                        return
+
+            if model_config.state_config is not None:
+                processor = CustomStateClassificationProcessor(
+                    self.config, model_config, self.requestor, self.metrics
+                )
+            else:
+                processor = CustomObjectClassificationProcessor(
+                    self.config,
+                    model_config,
+                    self.event_metadata_publisher,
+                    self.metrics,
+                )
+
+            self.realtime_processors.append(processor)
+            logger.info(
+                f"Added classification processor for model: {model_name} (type: {type(processor).__name__})"
+            )
+
     def _process_requests(self) -> None:
         """Process embeddings requests"""
 

frigate/util/classification.py

@@ -2,12 +2,15 @@
 
 import logging
 import os
+import random
+from collections import defaultdict
 
 import cv2
 import numpy as np
 
 from frigate.comms.embeddings_updater import EmbeddingsRequestEnum, EmbeddingsRequestor
 from frigate.comms.inter_process import InterProcessRequestor
+from frigate.config import FfmpegConfig
 from frigate.const import (
     CLIPS_DIR,
     MODEL_CACHE_DIR,
@@ -15,7 +18,10 @@ from frigate.const import (
     UPDATE_MODEL_STATE,
 )
 from frigate.log import redirect_output_to_logger
+from frigate.models import Event, Recordings, ReviewSegment
 from frigate.types import ModelStatusTypesEnum
+from frigate.util.image import get_image_from_recording
+from frigate.util.path import get_event_thumbnail_bytes
 from frigate.util.process import FrigateProcess
 
 BATCH_SIZE = 16
@@ -69,6 +75,7 @@ class ClassificationTrainingProcess(FrigateProcess):
         logger.info(f"Kicking off classification training for {self.model_name}.")
         dataset_dir = os.path.join(CLIPS_DIR, self.model_name, "dataset")
         model_dir = os.path.join(MODEL_CACHE_DIR, self.model_name)
+        os.makedirs(model_dir, exist_ok=True)
         num_classes = len(
             [
                 d
@@ -139,7 +146,6 @@ class ClassificationTrainingProcess(FrigateProcess):
             f.write(tflite_model)
 
 
-@staticmethod
 def kickoff_model_training(
     embeddingRequestor: EmbeddingsRequestor, model_name: str
 ) -> None:
@@ -172,3 +178,520 @@ def kickoff_model_training(
         },
     )
     requestor.stop()
+
+
+@staticmethod
+def collect_state_classification_examples(
+    model_name: str, cameras: dict[str, tuple[float, float, float, float]]
+) -> None:
+    """
+    Collect representative state classification examples from review items.
+
+    This function:
+    1. Queries review items from specified cameras
+    2. Selects 100 balanced timestamps across the data
+    3. Extracts keyframes from recordings (cropped to specified regions)
+    4. Selects 20 most visually distinct images
+    5. Saves them to the dataset directory
+
+    Args:
+        model_name: Name of the classification model
+        cameras: Dict mapping camera names to normalized crop coordinates [x1, y1, x2, y2] (0-1)
+    """
+    dataset_dir = os.path.join(CLIPS_DIR, model_name, "dataset")
+    temp_dir = os.path.join(dataset_dir, "temp")
+    os.makedirs(temp_dir, exist_ok=True)
+
+    # Step 1: Get review items for the cameras
+    camera_names = list(cameras.keys())
+    review_items = list(
+        ReviewSegment.select()
+        .where(ReviewSegment.camera.in_(camera_names))
+        .where(ReviewSegment.end_time.is_null(False))
+        .order_by(ReviewSegment.start_time.asc())
+    )
+
+    if not review_items:
+        logger.warning(f"No review items found for cameras: {camera_names}")
+        return
+
+    # Step 2: Create balanced timestamp selection (100 samples)
+    timestamps = _select_balanced_timestamps(review_items, target_count=100)
+
+    # Step 3: Extract keyframes from recordings with crops applied
+    keyframes = _extract_keyframes(
+        "/usr/lib/ffmpeg/7.0/bin/ffmpeg", timestamps, temp_dir, cameras
+    )
+
+    # Step 4: Select 24 most visually distinct images (they're already cropped)
+    distinct_images = _select_distinct_images(keyframes, target_count=24)
+
+    # Step 5: Save to train directory for later classification
+    train_dir = os.path.join(CLIPS_DIR, model_name, "train")
+    os.makedirs(train_dir, exist_ok=True)
+
+    saved_count = 0
+    for idx, image_path in enumerate(distinct_images):
+        dest_path = os.path.join(train_dir, f"example_{idx:03d}.jpg")
+        try:
+            img = cv2.imread(image_path)
+
+            if img is not None:
+                cv2.imwrite(dest_path, img)
+                saved_count += 1
+        except Exception as e:
+            logger.error(f"Failed to save image {image_path}: {e}")
+
+    import shutil
+
+    try:
+        shutil.rmtree(temp_dir)
+    except Exception as e:
+        logger.warning(f"Failed to clean up temp directory: {e}")
+
+
+def _select_balanced_timestamps(
+    review_items: list[ReviewSegment], target_count: int = 100
+) -> list[dict]:
+    """
+    Select balanced timestamps from review items.
+
+    Strategy:
+    - Group review items by camera and time of day
+    - Sample evenly across groups to ensure diversity
+    - For each selected review item, pick a random timestamp within its duration
+
+    Returns:
+        List of dicts with keys: camera, timestamp, review_item
+    """
+    # Group by camera and hour of day for temporal diversity
+    grouped = defaultdict(list)
+
+    for item in review_items:
+        camera = item.camera
+        # Group by 6-hour blocks for temporal diversity
+        hour_block = int(item.start_time // (6 * 3600))
+        key = f"{camera}_{hour_block}"
+        grouped[key].append(item)
+
+    # Calculate how many samples per group
+    num_groups = len(grouped)
+    if num_groups == 0:
+        return []
+
+    samples_per_group = max(1, target_count // num_groups)
+    timestamps = []
+
+    # Sample from each group
+    for group_items in grouped.values():
+        # Take samples_per_group items from this group
+        sample_size = min(samples_per_group, len(group_items))
+        sampled_items = random.sample(group_items, sample_size)
+
+        for item in sampled_items:
+            # Pick a random timestamp within the review item's duration
+            duration = item.end_time - item.start_time
+            if duration <= 0:
+                continue
+
+            # Sample from middle 80% to avoid edge artifacts
+            offset = random.uniform(duration * 0.1, duration * 0.9)
+            timestamp = item.start_time + offset
+
+            timestamps.append(
+                {
+                    "camera": item.camera,
+                    "timestamp": timestamp,
+                    "review_item": item,
+                }
+            )
+
+    # If we don't have enough, sample more from larger groups
+    while len(timestamps) < target_count and len(timestamps) < len(review_items):
+        for group_items in grouped.values():
+            if len(timestamps) >= target_count:
+                break
+
+            # Pick a random item not already sampled
+            item = random.choice(group_items)
+            duration = item.end_time - item.start_time
+            if duration <= 0:
+                continue
+
+            offset = random.uniform(duration * 0.1, duration * 0.9)
+            timestamp = item.start_time + offset
+
+            # Check if we already have a timestamp near this one
+            if not any(abs(t["timestamp"] - timestamp) < 1.0 for t in timestamps):
+                timestamps.append(
+                    {
+                        "camera": item.camera,
+                        "timestamp": timestamp,
+                        "review_item": item,
+                    }
+                )
+
+    return timestamps[:target_count]
+
+
+def _extract_keyframes(
+    ffmpeg_path: str,
+    timestamps: list[dict],
+    output_dir: str,
+    camera_crops: dict[str, tuple[float, float, float, float]],
+) -> list[str]:
+    """
+    Extract keyframes from recordings at specified timestamps and crop to specified regions.
+
+    Args:
+        ffmpeg_path: Path to ffmpeg binary
+        timestamps: List of timestamp dicts from _select_balanced_timestamps
+        output_dir: Directory to save extracted frames
+        camera_crops: Dict mapping camera names to normalized crop coordinates [x1, y1, x2, y2] (0-1)
+
+    Returns:
+        List of paths to successfully extracted and cropped keyframe images
+    """
+    keyframe_paths = []
+
+    for idx, ts_info in enumerate(timestamps):
+        camera = ts_info["camera"]
+        timestamp = ts_info["timestamp"]
+
+        if camera not in camera_crops:
+            logger.warning(f"No crop coordinates for camera {camera}")
+            continue
+
+        norm_x1, norm_y1, norm_x2, norm_y2 = camera_crops[camera]
+
+        try:
+            recording = (
+                Recordings.select()
+                .where(
+                    (timestamp >= Recordings.start_time)
+                    & (timestamp <= Recordings.end_time)
+                    & (Recordings.camera == camera)
+                )
+                .order_by(Recordings.start_time.desc())
+                .limit(1)
+                .get()
+            )
+        except Exception:
+            continue
+
+        relative_time = timestamp - recording.start_time
+
+        try:
+            config = FfmpegConfig(path="/usr/lib/ffmpeg/7.0")
+            image_data = get_image_from_recording(
+                config,
+                recording.path,
+                relative_time,
+                codec="mjpeg",
+                height=None,
+            )
+
+            if image_data:
+                nparr = np.frombuffer(image_data, np.uint8)
+                img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
+
+                if img is not None:
+                    height, width = img.shape[:2]
+
+                    x1 = int(norm_x1 * width)
+                    y1 = int(norm_y1 * height)
+                    x2 = int(norm_x2 * width)
+                    y2 = int(norm_y2 * height)
+
+                    x1_clipped = max(0, min(x1, width))
+                    y1_clipped = max(0, min(y1, height))
+                    x2_clipped = max(0, min(x2, width))
+                    y2_clipped = max(0, min(y2, height))
+
+                    if x2_clipped > x1_clipped and y2_clipped > y1_clipped:
+                        cropped = img[y1_clipped:y2_clipped, x1_clipped:x2_clipped]
+                        resized = cv2.resize(cropped, (224, 224))
+
+                        output_path = os.path.join(output_dir, f"frame_{idx:04d}.jpg")
+                        cv2.imwrite(output_path, resized)
+                        keyframe_paths.append(output_path)
+
+        except Exception as e:
+            logger.debug(
+                f"Failed to extract frame from {recording.path} at {relative_time}s: {e}"
+            )
+            continue
+
+    return keyframe_paths
+
+
+def _select_distinct_images(
+    image_paths: list[str], target_count: int = 20
+) -> list[str]:
+    """
+    Select the most visually distinct images from a set of keyframes.
+
+    Uses a greedy algorithm based on image histograms:
+    1. Start with a random image
+    2. Iteratively add the image that is most different from already selected images
+    3. Difference is measured using histogram comparison
+
+    Args:
+        image_paths: List of paths to candidate images
+        target_count: Number of distinct images to select
+
+    Returns:
+        List of paths to selected images
+    """
+    if len(image_paths) <= target_count:
+        return image_paths
+
+    histograms = {}
+    valid_paths = []
+
+    for path in image_paths:
+        try:
+            img = cv2.imread(path)
+
+            if img is None:
+                continue
+
+            hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
+            hist = cv2.calcHist(
+                [hsv], [0, 1, 2], None, [8, 8, 8], [0, 180, 0, 256, 0, 256]
+            )
+            hist = cv2.normalize(hist, hist).flatten()
+            histograms[path] = hist
+            valid_paths.append(path)
+        except Exception as e:
+            logger.debug(f"Failed to process image {path}: {e}")
+            continue
+
+    if len(valid_paths) <= target_count:
+        return valid_paths
+
+    selected = []
+    first_image = random.choice(valid_paths)
+    selected.append(first_image)
+    remaining = [p for p in valid_paths if p != first_image]
+
+    while len(selected) < target_count and remaining:
+        max_min_distance = -1
+        best_candidate = None
+
+        for candidate in remaining:
+            min_distance = float("inf")
+
+            for selected_img in selected:
+                distance = cv2.compareHist(
+                    histograms[candidate],
+                    histograms[selected_img],
+                    cv2.HISTCMP_BHATTACHARYYA,
+                )
+                min_distance = min(min_distance, distance)
+
+            if min_distance > max_min_distance:
+                max_min_distance = min_distance
+                best_candidate = candidate
+
+        if best_candidate:
+            selected.append(best_candidate)
+            remaining.remove(best_candidate)
+        else:
+            break
+
+    return selected
+
+
+@staticmethod
+def collect_object_classification_examples(
+    model_name: str,
+    label: str,
+) -> None:
+    """
+    Collect representative object classification examples from event thumbnails.
+
+    This function:
+    1. Queries events for the specified label
+    2. Selects 100 balanced events across different cameras and times
+    3. Retrieves thumbnails for selected events (with 33% center crop applied)
+    4. Selects 24 most visually distinct thumbnails
+    5. Saves to dataset directory
+
+    Args:
+        model_name: Name of the classification model
+        label: Object label to collect (e.g., "person", "car")
+        cameras: List of camera names to collect examples from
+    """
+    dataset_dir = os.path.join(CLIPS_DIR, model_name, "dataset")
+    temp_dir = os.path.join(dataset_dir, "temp")
+    os.makedirs(temp_dir, exist_ok=True)
+
+    # Step 1: Query events for the specified label and cameras
+    events = list(
+        Event.select().where((Event.label == label)).order_by(Event.start_time.asc())
+    )
+
+    if not events:
+        logger.warning(f"No events found for label '{label}'")
+        return
+
+    logger.debug(f"Found {len(events)} events")
+
+    # Step 2: Select balanced events (100 samples)
+    selected_events = _select_balanced_events(events, target_count=100)
+    logger.debug(f"Selected {len(selected_events)} events")
+
+    # Step 3: Extract thumbnails from events
+    thumbnails = _extract_event_thumbnails(selected_events, temp_dir)
+    logger.debug(f"Successfully extracted {len(thumbnails)} thumbnails")
+
+    # Step 4: Select 24 most visually distinct thumbnails
+    distinct_images = _select_distinct_images(thumbnails, target_count=24)
+    logger.debug(f"Selected {len(distinct_images)} distinct images")
+
+    # Step 5: Save to train directory for later classification
+    train_dir = os.path.join(CLIPS_DIR, model_name, "train")
+    os.makedirs(train_dir, exist_ok=True)
+
+    saved_count = 0
+    for idx, image_path in enumerate(distinct_images):
+        dest_path = os.path.join(train_dir, f"example_{idx:03d}.jpg")
+        try:
+            img = cv2.imread(image_path)
+
+            if img is not None:
+                cv2.imwrite(dest_path, img)
+                saved_count += 1
+        except Exception as e:
+            logger.error(f"Failed to save image {image_path}: {e}")
+
+    import shutil
+
+    try:
+        shutil.rmtree(temp_dir)
+    except Exception as e:
+        logger.warning(f"Failed to clean up temp directory: {e}")
+
+    logger.debug(
+        f"Successfully collected {saved_count} classification examples in {train_dir}"
+    )
+
+
+def _select_balanced_events(
+    events: list[Event], target_count: int = 100
+) -> list[Event]:
+    """
+    Select balanced events from the event list.
+
+    Strategy:
+    - Group events by camera and time of day
+    - Sample evenly across groups to ensure diversity
+    - Prioritize events with higher scores
+
+    Returns:
+        List of selected events
+    """
+    grouped = defaultdict(list)
+
+    for event in events:
+        camera = event.camera
+        hour_block = int(event.start_time // (6 * 3600))
+        key = f"{camera}_{hour_block}"
+        grouped[key].append(event)
+
+    num_groups = len(grouped)
+    if num_groups == 0:
+        return []
+
+    samples_per_group = max(1, target_count // num_groups)
+    selected = []
+
+    for group_events in grouped.values():
+        sorted_events = sorted(
+            group_events,
+            key=lambda e: e.data.get("score", 0) if e.data else 0,
+            reverse=True,
+        )
+
+        sample_size = min(samples_per_group, len(sorted_events))
+        selected.extend(sorted_events[:sample_size])
+
+    if len(selected) < target_count:
+        remaining = [e for e in events if e not in selected]
+        remaining_sorted = sorted(
+            remaining,
+            key=lambda e: e.data.get("score", 0) if e.data else 0,
+            reverse=True,
+        )
+        needed = target_count - len(selected)
+        selected.extend(remaining_sorted[:needed])
+
+    return selected[:target_count]
+
+
+def _extract_event_thumbnails(events: list[Event], output_dir: str) -> list[str]:
+    """
+    Extract thumbnails from events and save to disk.
+
+    Args:
+        events: List of Event objects
+        output_dir: Directory to save thumbnails
+
+    Returns:
+        List of paths to successfully extracted thumbnail images
+    """
+    thumbnail_paths = []
+
+    for idx, event in enumerate(events):
+        try:
+            thumbnail_bytes = get_event_thumbnail_bytes(event)
+
+            if thumbnail_bytes:
+                nparr = np.frombuffer(thumbnail_bytes, np.uint8)
+                img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
+
+                if img is not None:
+                    height, width = img.shape[:2]
+
+                    crop_size = 1.0
+                    if event.data and "box" in event.data and "region" in event.data:
+                        box = event.data["box"]
+                        region = event.data["region"]
+
+                        if len(box) == 4 and len(region) == 4:
+                            box_w, box_h = box[2], box[3]
+                            region_w, region_h = region[2], region[3]
+
+                            box_area = (box_w * box_h) / (region_w * region_h)
+
+                            if box_area < 0.05:
+                                crop_size = 0.4
+                            elif box_area < 0.10:
+                                crop_size = 0.5
+                            elif box_area < 0.20:
+                                crop_size = 0.65
+                            elif box_area < 0.35:
+                                crop_size = 0.80
+                            else:
+                                crop_size = 0.95
+
+                    crop_width = int(width * crop_size)
+                    crop_height = int(height * crop_size)
+
+                    x1 = (width - crop_width) // 2
+                    y1 = (height - crop_height) // 2
+                    x2 = x1 + crop_width
+                    y2 = y1 + crop_height
+
+                    cropped = img[y1:y2, x1:x2]
+                    resized = cv2.resize(cropped, (224, 224))
+                    output_path = os.path.join(output_dir, f"thumbnail_{idx:04d}.jpg")
+                    cv2.imwrite(output_path, resized)
+                    thumbnail_paths.append(output_path)
+
+        except Exception as e:
+            logger.debug(f"Failed to extract thumbnail for event {event.id}: {e}")
+            continue
+
+    return thumbnail_paths

web/public/locales/en/views/classificationModel.json

@@ -1,4 +1,5 @@
 {
+  "documentTitle": "Classification Models",
   "button": {
     "deleteClassificationAttempts": "Delete Classification Images",
     "renameCategory": "Rename Class",
@@ -50,8 +51,85 @@
   },
   "categorizeImageAs": "Classify Image As:",
   "categorizeImage": "Classify Image",
+  "noModels": {
+    "object": {
+      "title": "No Object Classification Models",
+      "description": "Create a custom model to classify detected objects.",
+      "buttonText": "Create Object Model"
+    },
+    "state": {
+      "title": "No State Classification Models",
+      "description": "Create a custom model to monitor and classify state changes in specific camera areas.",
+      "buttonText": "Create State Model"
+    }
+  },
   "wizard": {
     "title": "Create New Classification",
-    "description": "Create a new state or object classification model."
+    "steps": {
+      "nameAndDefine": "Name & Define",
+      "stateArea": "State Area",
+      "chooseExamples": "Choose Examples"
+    },
+    "step1": {
+      "description": "State models monitor fixed camera areas for changes (e.g., door open/closed). Object models add classifications to detected objects (e.g., known animals, delivery persons, etc.).",
+      "name": "Name",
+      "namePlaceholder": "Enter model name...",
+      "type": "Type",
+      "typeState": "State",
+      "typeObject": "Object",
+      "objectLabel": "Object Label",
+      "objectLabelPlaceholder": "Select object type...",
+      "classificationType": "Classification Type",
+      "classificationTypeTip": "Learn about classification types",
+      "classificationTypeDesc": "Sub Labels add additional text to the object label (e.g., 'Person: UPS'). Attributes are searchable metadata stored separately in the object metadata.",
+      "classificationSubLabel": "Sub Label",
+      "classificationAttribute": "Attribute",
+      "classes": "Classes",
+      "classesTip": "Learn about classes",
+      "classesStateDesc": "Define the different states your camera area can be in. For example: 'open' and 'closed' for a garage door.",
+      "classesObjectDesc": "Define the different categories to classify detected objects into. For example: 'delivery_person', 'resident', 'stranger' for person classification.",
+      "classPlaceholder": "Enter class name...",
+      "errors": {
+        "nameRequired": "Model name is required",
+        "nameLength": "Model name must be 64 characters or less",
+        "nameOnlyNumbers": "Model name cannot contain only numbers",
+        "classRequired": "At least 1 class is required",
+        "classesUnique": "Class names must be unique",
+        "stateRequiresTwoClasses": "State models require at least 2 classes",
+        "objectLabelRequired": "Please select an object label",
+        "objectTypeRequired": "Please select a classification type"
+      }
+    },
+    "step2": {
+      "description": "Select cameras and define the area to monitor for each camera. The model will classify the state of these areas.",
+      "cameras": "Cameras",
+      "selectCamera": "Select Camera",
+      "noCameras": "Click + to add cameras",
+      "selectCameraPrompt": "Select a camera from the list to define its monitoring area"
+    },
+    "step3": {
+      "selectImagesPrompt": "Select all images with: {{className}}",
+      "selectImagesDescription": "Click on images to select them. Click Continue when you're done with this class.",
+      "generating": {
+        "title": "Generating Sample Images",
+        "description": "Frigate is pulling representative images from your recordings. This may take a moment..."
+      },
+      "training": {
+        "title": "Training Model",
+        "description": "Your model is being trained in the background. Close this dialog, and your model will start running as soon as training is complete."
+      },
+      "retryGenerate": "Retry Generation",
+      "noImages": "No sample images generated",
+      "classifying": "Classifying & Training...",
+      "trainingStarted": "Training started successfully",
+      "errors": {
+        "noCameras": "No cameras configured",
+        "noObjectLabel": "No object label selected",
+        "generateFailed": "Failed to generate examples: {{error}}",
+        "generationFailed": "Generation failed. Please try again.",
+        "classifyFailed": "Failed to classify images: {{error}}"
+      },
+      "generateSuccess": "Successfully generated sample images"
+    }
   }
 }

web/public/locales/en/views/faceLibrary.json

@@ -5,10 +5,6 @@
     "invalidName": "Invalid name. Names can only include letters, numbers, spaces, apostrophes, underscores, and hyphens."
   },
   "details": {
-    "subLabelScore": "Sub Label Score",
-    "scoreInfo": "The sub label score is the weighted score for all of the recognized face confidences, so this may differ from the score shown on the snapshot.",
-    "face": "Face Details",
-    "faceDesc": "Details of the tracked object that generated this face",
     "timestamp": "Timestamp",
     "unknown": "Unknown"
   },
@@ -19,8 +15,6 @@
   },
   "collections": "Collections",
   "createFaceLibrary": {
-    "title": "Create Collection",
-    "desc": "Create a new collection",
     "new": "Create New Face",
     "nextSteps": "To build a strong foundation:<li>Use the Recent Recognitions tab to select and train on images for each detected person.</li><li>Focus on straight-on images for best results; avoid training images that capture faces at an angle.</li></ul>"
   },
@@ -37,8 +31,6 @@
     "aria": "Select recent recognitions",
     "empty": "There are no recent face recognition attempts"
   },
-  "selectItem": "Select {{item}}",
-  "selectFace": "Select Face",
   "deleteFaceLibrary": {
     "title": "Delete Name",
     "desc": "Are you sure you want to delete the collection {{name}}? This will permanently delete all associated faces."
@@ -69,7 +61,6 @@
     "maxSize": "Max size: {{size}}MB"
   },
   "nofaces": "No faces available",
-  "pixels": "{{area}}px",
   "trainFaceAs": "Train Face as:",
   "trainFace": "Train Face",
   "toast": {

web/src/components/card/ClassificationCard.tsx

@@ -126,6 +126,7 @@ export const ClassificationCard = forwardRef<
           imgClassName,
           isMobile && "w-full",
         )}
+        loading="lazy"
         onLoad={() => setImageLoaded(true)}
         src={`${baseUrl}${data.filepath}`}
       />

web/src/components/classification/ClassificationModelWizardDialog.tsx

@@ -7,58 +7,198 @@ import {
   DialogHeader,
   DialogTitle,
 } from "../ui/dialog";
-import { useState } from "react";
+import { useReducer, useMemo } from "react";
+import Step1NameAndDefine, { Step1FormData } from "./wizard/Step1NameAndDefine";
+import Step2StateArea, { Step2FormData } from "./wizard/Step2StateArea";
+import Step3ChooseExamples, {
+  Step3FormData,
+} from "./wizard/Step3ChooseExamples";
+import { cn } from "@/lib/utils";
+import { isDesktop } from "react-device-detect";
 
-const STEPS = [
+const OBJECT_STEPS = [
-  "classificationWizard.steps.nameAndDefine",
+  "wizard.steps.nameAndDefine",
-  "classificationWizard.steps.stateArea",
+  "wizard.steps.chooseExamples",
-  "classificationWizard.steps.chooseExamples",
+];
-  "classificationWizard.steps.train",
+
+const STATE_STEPS = [
+  "wizard.steps.nameAndDefine",
+  "wizard.steps.stateArea",
+  "wizard.steps.chooseExamples",
 ];
 
 type ClassificationModelWizardDialogProps = {
   open: boolean;
   onClose: () => void;
+  defaultModelType?: "state" | "object";
 };
+
+type WizardState = {
+  currentStep: number;
+  step1Data?: Step1FormData;
+  step2Data?: Step2FormData;
+  step3Data?: Step3FormData;
+};
+
+type WizardAction =
+  | { type: "NEXT_STEP"; payload?: Partial<WizardState> }
+  | { type: "PREVIOUS_STEP" }
+  | { type: "SET_STEP_1"; payload: Step1FormData }
+  | { type: "SET_STEP_2"; payload: Step2FormData }
+  | { type: "SET_STEP_3"; payload: Step3FormData }
+  | { type: "RESET" };
+
+const initialState: WizardState = {
+  currentStep: 0,
+};
+
+function wizardReducer(state: WizardState, action: WizardAction): WizardState {
+  switch (action.type) {
+    case "SET_STEP_1":
+      return {
+        ...state,
+        step1Data: action.payload,
+        currentStep: 1,
+      };
+    case "SET_STEP_2":
+      return {
+        ...state,
+        step2Data: action.payload,
+        currentStep: 2,
+      };
+    case "SET_STEP_3":
+      return {
+        ...state,
+        step3Data: action.payload,
+        currentStep: 3,
+      };
+    case "NEXT_STEP":
+      return {
+        ...state,
+        ...action.payload,
+        currentStep: state.currentStep + 1,
+      };
+    case "PREVIOUS_STEP":
+      return {
+        ...state,
+        currentStep: Math.max(0, state.currentStep - 1),
+      };
+    case "RESET":
+      return initialState;
+    default:
+      return state;
+  }
+}
+
 export default function ClassificationModelWizardDialog({
   open,
   onClose,
+  defaultModelType,
 }: ClassificationModelWizardDialogProps) {
   const { t } = useTranslation(["views/classificationModel"]);
 
-  // step management
+  const [wizardState, dispatch] = useReducer(wizardReducer, initialState);
-  const [currentStep, _] = useState(0);
+
+  const steps = useMemo(() => {
+    if (!wizardState.step1Data) {
+      return OBJECT_STEPS;
+    }
+    return wizardState.step1Data.modelType === "state"
+      ? STATE_STEPS
+      : OBJECT_STEPS;
+  }, [wizardState.step1Data]);
+
+  const handleStep1Next = (data: Step1FormData) => {
+    dispatch({ type: "SET_STEP_1", payload: data });
+  };
+
+  const handleStep2Next = (data: Step2FormData) => {
+    dispatch({ type: "SET_STEP_2", payload: data });
+  };
+
+  const handleBack = () => {
+    dispatch({ type: "PREVIOUS_STEP" });
+  };
+
+  const handleCancel = () => {
+    dispatch({ type: "RESET" });
+    onClose();
+  };
 
   return (
     <Dialog
       open={open}
       onOpenChange={(open) => {
         if (!open) {
-          onClose;
+          handleCancel();
         }
       }}
     >
       <DialogContent
-        className="max-h-[90dvh] max-w-4xl overflow-y-auto"
+        className={cn(
+          "",
+          isDesktop &&
+            wizardState.currentStep == 0 &&
+            "max-h-[90%] overflow-y-auto xl:max-h-[80%]",
+          isDesktop &&
+            wizardState.currentStep > 0 &&
+            "max-h-[90%] max-w-[70%] overflow-y-auto xl:max-h-[80%]",
+        )}
         onInteractOutside={(e) => {
           e.preventDefault();
         }}
       >
         <StepIndicator
-          steps={STEPS}
+          steps={steps}
-          currentStep={currentStep}
+          currentStep={wizardState.currentStep}
           variant="dots"
           className="mb-4 justify-start"
         />
         <DialogHeader>
           <DialogTitle>{t("wizard.title")}</DialogTitle>
-          {currentStep === 0 && (
+          {wizardState.currentStep === 0 && (
-            <DialogDescription>{t("wizard.description")}</DialogDescription>
+            <DialogDescription>
+              {t("wizard.step1.description")}
+            </DialogDescription>
+          )}
+          {wizardState.currentStep === 1 &&
+            wizardState.step1Data?.modelType === "state" && (
+              <DialogDescription>
+                {t("wizard.step2.description")}
+              </DialogDescription>
             )}
         </DialogHeader>
 
         <div className="pb-4">
-          <div className="size-full"></div>
+          {wizardState.currentStep === 0 && (
+            <Step1NameAndDefine
+              initialData={wizardState.step1Data}
+              defaultModelType={defaultModelType}
+              onNext={handleStep1Next}
+              onCancel={handleCancel}
+            />
+          )}
+          {wizardState.currentStep === 1 &&
+            wizardState.step1Data?.modelType === "state" && (
+              <Step2StateArea
+                initialData={wizardState.step2Data}
+                onNext={handleStep2Next}
+                onBack={handleBack}
+              />
+            )}
+          {((wizardState.currentStep === 2 &&
+            wizardState.step1Data?.modelType === "state") ||
+            (wizardState.currentStep === 1 &&
+              wizardState.step1Data?.modelType === "object")) &&
+            wizardState.step1Data && (
+              <Step3ChooseExamples
+                step1Data={wizardState.step1Data}
+                step2Data={wizardState.step2Data}
+                initialData={wizardState.step3Data}
+                onClose={onClose}
+                onBack={handleBack}
+              />
+            )}
         </div>
       </DialogContent>
     </Dialog>

web/src/components/classification/wizard/Step1NameAndDefine.tsx

@@ -0,0 +1,498 @@
+import { Button } from "@/components/ui/button";
+import {
+  Form,
+  FormControl,
+  FormField,
+  FormItem,
+  FormLabel,
+  FormMessage,
+} from "@/components/ui/form";
+import { Input } from "@/components/ui/input";
+import { Label } from "@/components/ui/label";
+import { RadioGroup, RadioGroupItem } from "@/components/ui/radio-group";
+import {
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectTrigger,
+  SelectValue,
+} from "@/components/ui/select";
+import { useForm } from "react-hook-form";
+import { zodResolver } from "@hookform/resolvers/zod";
+import { z } from "zod";
+import { useTranslation } from "react-i18next";
+import { useMemo } from "react";
+import { LuX, LuPlus, LuInfo, LuExternalLink } from "react-icons/lu";
+import useSWR from "swr";
+import { FrigateConfig } from "@/types/frigateConfig";
+import { getTranslatedLabel } from "@/utils/i18n";
+import { useDocDomain } from "@/hooks/use-doc-domain";
+import {
+  Popover,
+  PopoverContent,
+  PopoverTrigger,
+} from "@/components/ui/popover";
+
+export type ModelType = "state" | "object";
+export type ObjectClassificationType = "sub_label" | "attribute";
+
+export type Step1FormData = {
+  modelName: string;
+  modelType: ModelType;
+  objectLabel?: string;
+  objectType?: ObjectClassificationType;
+  classes: string[];
+};
+
+type Step1NameAndDefineProps = {
+  initialData?: Partial<Step1FormData>;
+  defaultModelType?: "state" | "object";
+  onNext: (data: Step1FormData) => void;
+  onCancel: () => void;
+};
+
+export default function Step1NameAndDefine({
+  initialData,
+  defaultModelType,
+  onNext,
+  onCancel,
+}: Step1NameAndDefineProps) {
+  const { t } = useTranslation(["views/classificationModel"]);
+  const { data: config } = useSWR<FrigateConfig>("config");
+  const { getLocaleDocUrl } = useDocDomain();
+
+  const objectLabels = useMemo(() => {
+    if (!config) return [];
+
+    const labels = new Set<string>();
+
+    Object.values(config.cameras).forEach((cameraConfig) => {
+      if (!cameraConfig.enabled || !cameraConfig.enabled_in_config) {
+        return;
+      }
+
+      cameraConfig.objects.track.forEach((label) => {
+        if (!config.model.all_attributes.includes(label)) {
+          labels.add(label);
+        }
+      });
+    });
+
+    return [...labels].sort();
+  }, [config]);
+
+  const step1FormData = z
+    .object({
+      modelName: z
+        .string()
+        .min(1, t("wizard.step1.errors.nameRequired"))
+        .max(64, t("wizard.step1.errors.nameLength"))
+        .refine((value) => !/^\d+$/.test(value), {
+          message: t("wizard.step1.errors.nameOnlyNumbers"),
+        }),
+      modelType: z.enum(["state", "object"]),
+      objectLabel: z.string().optional(),
+      objectType: z.enum(["sub_label", "attribute"]).optional(),
+      classes: z
+        .array(z.string())
+        .min(1, t("wizard.step1.errors.classRequired"))
+        .refine(
+          (classes) => {
+            const nonEmpty = classes.filter((c) => c.trim().length > 0);
+            return nonEmpty.length >= 1;
+          },
+          { message: t("wizard.step1.errors.classRequired") },
+        )
+        .refine(
+          (classes) => {
+            const nonEmpty = classes.filter((c) => c.trim().length > 0);
+            const unique = new Set(nonEmpty.map((c) => c.toLowerCase()));
+            return unique.size === nonEmpty.length;
+          },
+          { message: t("wizard.step1.errors.classesUnique") },
+        ),
+    })
+    .refine(
+      (data) => {
+        // State models require at least 2 classes
+        if (data.modelType === "state") {
+          const nonEmpty = data.classes.filter((c) => c.trim().length > 0);
+          return nonEmpty.length >= 2;
+        }
+        return true;
+      },
+      {
+        message: t("wizard.step1.errors.stateRequiresTwoClasses"),
+        path: ["classes"],
+      },
+    )
+    .refine(
+      (data) => {
+        if (data.modelType === "object") {
+          return data.objectLabel !== undefined && data.objectLabel !== "";
+        }
+        return true;
+      },
+      {
+        message: t("wizard.step1.errors.objectLabelRequired"),
+        path: ["objectLabel"],
+      },
+    )
+    .refine(
+      (data) => {
+        if (data.modelType === "object") {
+          return data.objectType !== undefined;
+        }
+        return true;
+      },
+      {
+        message: t("wizard.step1.errors.objectTypeRequired"),
+        path: ["objectType"],
+      },
+    );
+
+  const form = useForm<z.infer<typeof step1FormData>>({
+    resolver: zodResolver(step1FormData),
+    defaultValues: {
+      modelName: initialData?.modelName || "",
+      modelType: initialData?.modelType || defaultModelType || "state",
+      objectLabel: initialData?.objectLabel,
+      objectType: initialData?.objectType || "sub_label",
+      classes: initialData?.classes?.length ? initialData.classes : [""],
+    },
+    mode: "onChange",
+  });
+
+  const watchedClasses = form.watch("classes");
+  const watchedModelType = form.watch("modelType");
+  const watchedObjectType = form.watch("objectType");
+
+  const handleAddClass = () => {
+    const currentClasses = form.getValues("classes");
+    form.setValue("classes", [...currentClasses, ""], { shouldValidate: true });
+  };
+
+  const handleRemoveClass = (index: number) => {
+    const currentClasses = form.getValues("classes");
+    const newClasses = currentClasses.filter((_, i) => i !== index);
+
+    // Ensure at least one field remains (even if empty)
+    if (newClasses.length === 0) {
+      form.setValue("classes", [""], { shouldValidate: true });
+    } else {
+      form.setValue("classes", newClasses, { shouldValidate: true });
+    }
+  };
+
+  const onSubmit = (data: z.infer<typeof step1FormData>) => {
+    // Filter out empty classes
+    const filteredClasses = data.classes.filter((c) => c.trim().length > 0);
+    onNext({
+      ...data,
+      classes: filteredClasses,
+    });
+  };
+
+  return (
+    <div className="space-y-6">
+      <Form {...form}>
+        <form onSubmit={form.handleSubmit(onSubmit)} className="space-y-4">
+          <FormField
+            control={form.control}
+            name="modelName"
+            render={({ field }) => (
+              <FormItem>
+                <FormLabel className="text-primary-variant">
+                  {t("wizard.step1.name")}
+                </FormLabel>
+                <FormControl>
+                  <Input
+                    className="h-8"
+                    placeholder={t("wizard.step1.namePlaceholder")}
+                    {...field}
+                  />
+                </FormControl>
+                <FormMessage />
+              </FormItem>
+            )}
+          />
+
+          <FormField
+            control={form.control}
+            name="modelType"
+            render={({ field }) => (
+              <FormItem>
+                <FormLabel className="text-primary-variant">
+                  {t("wizard.step1.type")}
+                </FormLabel>
+                <FormControl>
+                  <RadioGroup
+                    onValueChange={field.onChange}
+                    defaultValue={field.value}
+                    className="flex flex-col gap-4 pt-2"
+                  >
+                    <div className="flex items-center gap-2">
+                      <RadioGroupItem
+                        className={
+                          watchedModelType === "state"
+                            ? "bg-selected from-selected/50 to-selected/90 text-selected"
+                            : "bg-secondary from-secondary/50 to-secondary/90 text-secondary"
+                        }
+                        id="state"
+                        value="state"
+                      />
+                      <Label className="cursor-pointer" htmlFor="state">
+                        {t("wizard.step1.typeState")}
+                      </Label>
+                    </div>
+                    <div className="flex items-center gap-2">
+                      <RadioGroupItem
+                        className={
+                          watchedModelType === "object"
+                            ? "bg-selected from-selected/50 to-selected/90 text-selected"
+                            : "bg-secondary from-secondary/50 to-secondary/90 text-secondary"
+                        }
+                        id="object"
+                        value="object"
+                      />
+                      <Label className="cursor-pointer" htmlFor="object">
+                        {t("wizard.step1.typeObject")}
+                      </Label>
+                    </div>
+                  </RadioGroup>
+                </FormControl>
+                <FormMessage />
+              </FormItem>
+            )}
+          />
+
+          {watchedModelType === "object" && (
+            <>
+              <FormField
+                control={form.control}
+                name="objectLabel"
+                render={({ field }) => (
+                  <FormItem>
+                    <FormLabel className="text-primary-variant">
+                      {t("wizard.step1.objectLabel")}
+                    </FormLabel>
+                    <Select
+                      onValueChange={field.onChange}
+                      defaultValue={field.value}
+                    >
+                      <FormControl>
+                        <SelectTrigger className="h-8">
+                          <SelectValue
+                            placeholder={t(
+                              "wizard.step1.objectLabelPlaceholder",
+                            )}
+                          />
+                        </SelectTrigger>
+                      </FormControl>
+                      <SelectContent>
+                        {objectLabels.map((label) => (
+                          <SelectItem
+                            key={label}
+                            value={label}
+                            className="cursor-pointer hover:bg-secondary-highlight"
+                          >
+                            {getTranslatedLabel(label)}
+                          </SelectItem>
+                        ))}
+                      </SelectContent>
+                    </Select>
+                    <FormMessage />
+                  </FormItem>
+                )}
+              />
+
+              <FormField
+                control={form.control}
+                name="objectType"
+                render={({ field }) => (
+                  <FormItem>
+                    <div className="flex items-center gap-1">
+                      <FormLabel className="text-primary-variant">
+                        {t("wizard.step1.classificationType")}
+                      </FormLabel>
+                      <Popover>
+                        <PopoverTrigger asChild>
+                          <Button
+                            variant="ghost"
+                            size="sm"
+                            className="h-4 w-4 p-0"
+                          >
+                            <LuInfo className="size-3" />
+                          </Button>
+                        </PopoverTrigger>
+                        <PopoverContent className="pointer-events-auto w-80 text-xs">
+                          <div className="flex flex-col gap-2">
+                            <div className="text-sm">
+                              {t("wizard.step1.classificationTypeDesc")}
+                            </div>
+                            <div className="mt-3 flex items-center text-primary">
+                              <a
+                                href={getLocaleDocUrl(
+                                  "configuration/custom_classification/object_classification#classification-type",
+                                )}
+                                target="_blank"
+                                rel="noopener noreferrer"
+                                className="inline cursor-pointer"
+                              >
+                                {t("readTheDocumentation", { ns: "common" })}
+                                <LuExternalLink className="ml-2 inline-flex size-3" />
+                              </a>
+                            </div>
+                          </div>
+                        </PopoverContent>
+                      </Popover>
+                    </div>
+                    <FormControl>
+                      <RadioGroup
+                        onValueChange={field.onChange}
+                        defaultValue={field.value}
+                        className="flex flex-col gap-4 pt-2"
+                      >
+                        <div className="flex items-center gap-2">
+                          <RadioGroupItem
+                            className={
+                              watchedObjectType === "sub_label"
+                                ? "bg-selected from-selected/50 to-selected/90 text-selected"
+                                : "bg-secondary from-secondary/50 to-secondary/90 text-secondary"
+                            }
+                            id="sub_label"
+                            value="sub_label"
+                          />
+                          <Label className="cursor-pointer" htmlFor="sub_label">
+                            {t("wizard.step1.classificationSubLabel")}
+                          </Label>
+                        </div>
+                        <div className="flex items-center gap-2">
+                          <RadioGroupItem
+                            className={
+                              watchedObjectType === "attribute"
+                                ? "bg-selected from-selected/50 to-selected/90 text-selected"
+                                : "bg-secondary from-secondary/50 to-secondary/90 text-secondary"
+                            }
+                            id="attribute"
+                            value="attribute"
+                          />
+                          <Label className="cursor-pointer" htmlFor="attribute">
+                            {t("wizard.step1.classificationAttribute")}
+                          </Label>
+                        </div>
+                      </RadioGroup>
+                    </FormControl>
+                    <FormMessage />
+                  </FormItem>
+                )}
+              />
+            </>
+          )}
+
+          <div className="space-y-2">
+            <div className="flex items-center justify-between">
+              <div className="flex items-center gap-1">
+                <FormLabel className="text-primary-variant">
+                  {t("wizard.step1.classes")}
+                </FormLabel>
+                <Popover>
+                  <PopoverTrigger asChild>
+                    <Button variant="ghost" size="sm" className="h-4 w-4 p-0">
+                      <LuInfo className="size-3" />
+                    </Button>
+                  </PopoverTrigger>
+                  <PopoverContent className="pointer-events-auto w-80 text-xs">
+                    <div className="flex flex-col gap-2">
+                      <div className="text-sm">
+                        {watchedModelType === "state"
+                          ? t("wizard.step1.classesStateDesc")
+                          : t("wizard.step1.classesObjectDesc")}
+                      </div>
+                      <div className="mt-3 flex items-center text-primary">
+                        <a
+                          href={getLocaleDocUrl(
+                            watchedModelType === "state"
+                              ? "configuration/custom_classification/state_classification"
+                              : "configuration/custom_classification/object_classification",
+                          )}
+                          target="_blank"
+                          rel="noopener noreferrer"
+                          className="inline cursor-pointer"
+                        >
+                          {t("readTheDocumentation", { ns: "common" })}
+                          <LuExternalLink className="ml-2 inline-flex size-3" />
+                        </a>
+                      </div>
+                    </div>
+                  </PopoverContent>
+                </Popover>
+              </div>
+              <Button
+                type="button"
+                variant="secondary"
+                className="size-6 rounded-md bg-secondary-foreground p-1 text-background"
+                onClick={handleAddClass}
+              >
+                <LuPlus />
+              </Button>
+            </div>
+            <div className="space-y-2">
+              {watchedClasses.map((_, index) => (
+                <FormField
+                  key={index}
+                  control={form.control}
+                  name={`classes.${index}`}
+                  render={({ field }) => (
+                    <FormItem>
+                      <FormControl>
+                        <div className="flex items-center gap-2">
+                          <Input
+                            className="h-8"
+                            placeholder={t("wizard.step1.classPlaceholder")}
+                            {...field}
+                          />
+                          {watchedClasses.length > 1 && (
+                            <Button
+                              type="button"
+                              variant="ghost"
+                              size="sm"
+                              className="h-8 w-8 p-0"
+                              onClick={() => handleRemoveClass(index)}
+                            >
+                              <LuX className="size-4" />
+                            </Button>
+                          )}
+                        </div>
+                      </FormControl>
+                    </FormItem>
+                  )}
+                />
+              ))}
+            </div>
+            {form.formState.errors.classes && (
+              <p className="text-sm font-medium text-destructive">
+                {form.formState.errors.classes.message}
+              </p>
+            )}
+          </div>
+        </form>
+      </Form>
+
+      <div className="flex flex-col gap-3 pt-3 sm:flex-row sm:justify-end sm:gap-4">
+        <Button type="button" onClick={onCancel} className="sm:flex-1">
+          {t("button.cancel", { ns: "common" })}
+        </Button>
+        <Button
+          type="button"
+          onClick={form.handleSubmit(onSubmit)}
+          variant="select"
+          className="flex items-center justify-center gap-2 sm:flex-1"
+          disabled={!form.formState.isValid}
+        >
+          {t("button.continue", { ns: "common" })}
+        </Button>
+      </div>
+    </div>
+  );
+}

web/src/components/classification/wizard/Step2StateArea.tsx

@@ -0,0 +1,479 @@
+import { Button } from "@/components/ui/button";
+import { useTranslation } from "react-i18next";
+import { useState, useMemo, useRef, useCallback, useEffect } from "react";
+import useSWR from "swr";
+import { FrigateConfig } from "@/types/frigateConfig";
+import {
+  Popover,
+  PopoverContent,
+  PopoverTrigger,
+} from "@/components/ui/popover";
+import { LuX, LuPlus } from "react-icons/lu";
+import { Stage, Layer, Rect, Transformer } from "react-konva";
+import Konva from "konva";
+import { useResizeObserver } from "@/hooks/resize-observer";
+import { useApiHost } from "@/api";
+import { resolveCameraName } from "@/hooks/use-camera-friendly-name";
+import Heading from "@/components/ui/heading";
+import { isMobile } from "react-device-detect";
+import { cn } from "@/lib/utils";
+
+export type CameraAreaConfig = {
+  camera: string;
+  crop: [number, number, number, number];
+};
+
+export type Step2FormData = {
+  cameraAreas: CameraAreaConfig[];
+};
+
+type Step2StateAreaProps = {
+  initialData?: Partial<Step2FormData>;
+  onNext: (data: Step2FormData) => void;
+  onBack: () => void;
+};
+
+export default function Step2StateArea({
+  initialData,
+  onNext,
+  onBack,
+}: Step2StateAreaProps) {
+  const { t } = useTranslation(["views/classificationModel"]);
+  const { data: config } = useSWR<FrigateConfig>("config");
+  const apiHost = useApiHost();
+
+  const [cameraAreas, setCameraAreas] = useState<CameraAreaConfig[]>(
+    initialData?.cameraAreas || [],
+  );
+  const [selectedCameraIndex, setSelectedCameraIndex] = useState<number>(0);
+  const [isPopoverOpen, setIsPopoverOpen] = useState(false);
+  const [imageLoaded, setImageLoaded] = useState(false);
+
+  const containerRef = useRef<HTMLDivElement>(null);
+  const imageRef = useRef<HTMLImageElement>(null);
+  const stageRef = useRef<Konva.Stage>(null);
+  const rectRef = useRef<Konva.Rect>(null);
+  const transformerRef = useRef<Konva.Transformer>(null);
+
+  const [{ width: containerWidth }] = useResizeObserver(containerRef);
+
+  const availableCameras = useMemo(() => {
+    if (!config) return [];
+
+    const selectedCameraNames = cameraAreas.map((ca) => ca.camera);
+    return Object.entries(config.cameras)
+      .sort()
+      .filter(
+        ([name, cam]) =>
+          cam.enabled &&
+          cam.enabled_in_config &&
+          !selectedCameraNames.includes(name),
+      )
+      .map(([name]) => ({
+        name,
+        displayName: resolveCameraName(config, name),
+      }));
+  }, [config, cameraAreas]);
+
+  const selectedCamera = useMemo(() => {
+    if (cameraAreas.length === 0) return null;
+    return cameraAreas[selectedCameraIndex];
+  }, [cameraAreas, selectedCameraIndex]);
+
+  const selectedCameraConfig = useMemo(() => {
+    if (!config || !selectedCamera) return null;
+    return config.cameras[selectedCamera.camera];
+  }, [config, selectedCamera]);
+
+  const imageSize = useMemo(() => {
+    if (!containerWidth || !selectedCameraConfig) {
+      return { width: 0, height: 0 };
+    }
+
+    const containerAspectRatio = 16 / 9;
+    const containerHeight = containerWidth / containerAspectRatio;
+
+    const cameraAspectRatio =
+      selectedCameraConfig.detect.width / selectedCameraConfig.detect.height;
+
+    // Fit camera within 16:9 container
+    let imageWidth, imageHeight;
+    if (cameraAspectRatio > containerAspectRatio) {
+      imageWidth = containerWidth;
+      imageHeight = imageWidth / cameraAspectRatio;
+    } else {
+      imageHeight = containerHeight;
+      imageWidth = imageHeight * cameraAspectRatio;
+    }
+
+    return { width: imageWidth, height: imageHeight };
+  }, [containerWidth, selectedCameraConfig]);
+
+  const handleAddCamera = useCallback(
+    (cameraName: string) => {
+      // Calculate a square crop in pixel space
+      const camera = config?.cameras[cameraName];
+      if (!camera) return;
+
+      const cameraAspect = camera.detect.width / camera.detect.height;
+      const cropSize = 0.3;
+      let x1, y1, x2, y2;
+
+      if (cameraAspect >= 1) {
+        const pixelSize = cropSize * camera.detect.height;
+        const normalizedWidth = pixelSize / camera.detect.width;
+        x1 = (1 - normalizedWidth) / 2;
+        y1 = (1 - cropSize) / 2;
+        x2 = x1 + normalizedWidth;
+        y2 = y1 + cropSize;
+      } else {
+        const pixelSize = cropSize * camera.detect.width;
+        const normalizedHeight = pixelSize / camera.detect.height;
+        x1 = (1 - cropSize) / 2;
+        y1 = (1 - normalizedHeight) / 2;
+        x2 = x1 + cropSize;
+        y2 = y1 + normalizedHeight;
+      }
+
+      const newArea: CameraAreaConfig = {
+        camera: cameraName,
+        crop: [x1, y1, x2, y2],
+      };
+      setCameraAreas([...cameraAreas, newArea]);
+      setSelectedCameraIndex(cameraAreas.length);
+      setIsPopoverOpen(false);
+    },
+    [cameraAreas, config],
+  );
+
+  const handleRemoveCamera = useCallback(
+    (index: number) => {
+      const newAreas = cameraAreas.filter((_, i) => i !== index);
+      setCameraAreas(newAreas);
+      if (selectedCameraIndex >= newAreas.length) {
+        setSelectedCameraIndex(Math.max(0, newAreas.length - 1));
+      }
+    },
+    [cameraAreas, selectedCameraIndex],
+  );
+
+  const handleCropChange = useCallback(
+    (crop: [number, number, number, number]) => {
+      const newAreas = [...cameraAreas];
+      newAreas[selectedCameraIndex] = {
+        ...newAreas[selectedCameraIndex],
+        crop,
+      };
+      setCameraAreas(newAreas);
+    },
+    [cameraAreas, selectedCameraIndex],
+  );
+
+  useEffect(() => {
+    setImageLoaded(false);
+  }, [selectedCamera]);
+
+  useEffect(() => {
+    const rect = rectRef.current;
+    const transformer = transformerRef.current;
+
+    if (
+      rect &&
+      transformer &&
+      selectedCamera &&
+      imageSize.width > 0 &&
+      imageLoaded
+    ) {
+      rect.scaleX(1);
+      rect.scaleY(1);
+      transformer.nodes([rect]);
+      transformer.getLayer()?.batchDraw();
+    }
+  }, [selectedCamera, imageSize, imageLoaded]);
+
+  const handleRectChange = useCallback(() => {
+    const rect = rectRef.current;
+
+    if (rect && imageSize.width > 0) {
+      const actualWidth = rect.width() * rect.scaleX();
+      const actualHeight = rect.height() * rect.scaleY();
+
+      // Average dimensions to maintain perfect square
+      const size = (actualWidth + actualHeight) / 2;
+
+      rect.width(size);
+      rect.height(size);
+      rect.scaleX(1);
+      rect.scaleY(1);
+
+      const x1 = rect.x() / imageSize.width;
+      const y1 = rect.y() / imageSize.height;
+      const x2 = (rect.x() + size) / imageSize.width;
+      const y2 = (rect.y() + size) / imageSize.height;
+
+      handleCropChange([x1, y1, x2, y2]);
+    }
+  }, [imageSize, handleCropChange]);
+
+  const handleContinue = useCallback(() => {
+    onNext({ cameraAreas });
+  }, [cameraAreas, onNext]);
+
+  const canContinue = cameraAreas.length > 0;
+
+  return (
+    <div className="flex flex-col gap-4">
+      <div
+        className={cn(
+          "flex gap-4 overflow-hidden",
+          isMobile ? "flex-col" : "flex-row",
+        )}
+      >
+        <div
+          className={cn(
+            "flex flex-shrink-0 flex-col gap-2 overflow-y-auto rounded-lg bg-secondary p-4",
+            isMobile ? "w-full" : "w-64",
+          )}
+        >
+          <div className="flex items-center justify-between">
+            <h3 className="text-sm font-medium">{t("wizard.step2.cameras")}</h3>
+            {availableCameras.length > 0 ? (
+              <Popover
+                open={isPopoverOpen}
+                onOpenChange={setIsPopoverOpen}
+                modal={true}
+              >
+                <PopoverTrigger asChild>
+                  <Button
+                    type="button"
+                    variant="secondary"
+                    className="size-6 rounded-md bg-secondary-foreground p-1 text-background"
+                    aria-label="Add camera"
+                  >
+                    <LuPlus />
+                  </Button>
+                </PopoverTrigger>
+                <PopoverContent
+                  className="scrollbar-container w-64 border bg-background p-3 shadow-lg"
+                  align="start"
+                  sideOffset={5}
+                  onOpenAutoFocus={(e) => e.preventDefault()}
+                >
+                  <div className="flex flex-col gap-2">
+                    <Heading as="h4" className="text-sm text-primary-variant">
+                      {t("wizard.step2.selectCamera")}
+                    </Heading>
+                    <div className="scrollbar-container flex max-h-[30vh] flex-col gap-1 overflow-y-auto">
+                      {availableCameras.map((cam) => (
+                        <Button
+                          key={cam.name}
+                          type="button"
+                          variant="ghost"
+                          size="sm"
+                          className="h-auto justify-start p-2 capitalize text-primary"
+                          onClick={() => {
+                            handleAddCamera(cam.name);
+                          }}
+                        >
+                          {cam.displayName}
+                        </Button>
+                      ))}
+                    </div>
+                  </div>
+                </PopoverContent>
+              </Popover>
+            ) : (
+              <Button
+                variant="secondary"
+                className="size-6 cursor-not-allowed rounded-md bg-muted p-1 text-muted-foreground"
+                disabled
+              >
+                <LuPlus />
+              </Button>
+            )}
+          </div>
+
+          <div className="flex flex-col gap-1">
+            {cameraAreas.map((area, index) => {
+              const isSelected = index === selectedCameraIndex;
+              const displayName = resolveCameraName(config, area.camera);
+
+              return (
+                <div
+                  key={area.camera}
+                  className={`flex items-center justify-between rounded-md p-2 ${
+                    isSelected
+                      ? "bg-selected/20 ring-1 ring-selected"
+                      : "hover:bg-secondary/50"
+                  } cursor-pointer`}
+                  onClick={() => setSelectedCameraIndex(index)}
+                >
+                  <span className="text-sm capitalize">{displayName}</span>
+                  <Button
+                    type="button"
+                    variant="ghost"
+                    size="sm"
+                    className="h-6 w-6 p-0"
+                    onClick={(e) => {
+                      e.stopPropagation();
+                      handleRemoveCamera(index);
+                    }}
+                  >
+                    <LuX className="size-4" />
+                  </Button>
+                </div>
+              );
+            })}
+          </div>
+
+          {cameraAreas.length === 0 && (
+            <div className="flex flex-1 items-center justify-center text-center text-sm text-muted-foreground">
+              {t("wizard.step2.noCameras")}
+            </div>
+          )}
+        </div>
+
+        <div className="flex flex-1 items-center justify-center overflow-hidden rounded-lg p-4">
+          <div
+            ref={containerRef}
+            className="flex items-center justify-center"
+            style={{
+              width: "100%",
+              aspectRatio: "16 / 9",
+              maxHeight: "100%",
+            }}
+          >
+            {selectedCamera && selectedCameraConfig && imageSize.width > 0 ? (
+              <div
+                style={{
+                  width: imageSize.width,
+                  height: imageSize.height,
+                  position: "relative",
+                }}
+              >
+                <img
+                  ref={imageRef}
+                  src={`${apiHost}api/${selectedCamera.camera}/latest.jpg?h=500`}
+                  alt={resolveCameraName(config, selectedCamera.camera)}
+                  className="h-full w-full object-contain"
+                  onLoad={() => setImageLoaded(true)}
+                />
+                <Stage
+                  ref={stageRef}
+                  width={imageSize.width}
+                  height={imageSize.height}
+                  className="absolute inset-0"
+                >
+                  <Layer>
+                    <Rect
+                      ref={rectRef}
+                      x={selectedCamera.crop[0] * imageSize.width}
+                      y={selectedCamera.crop[1] * imageSize.height}
+                      width={
+                        (selectedCamera.crop[2] - selectedCamera.crop[0]) *
+                        imageSize.width
+                      }
+                      height={
+                        (selectedCamera.crop[3] - selectedCamera.crop[1]) *
+                        imageSize.height
+                      }
+                      stroke="#3b82f6"
+                      strokeWidth={2}
+                      fill="rgba(59, 130, 246, 0.1)"
+                      draggable
+                      dragBoundFunc={(pos) => {
+                        const rect = rectRef.current;
+                        if (!rect) return pos;
+
+                        const size = rect.width();
+                        const x = Math.max(
+                          0,
+                          Math.min(pos.x, imageSize.width - size),
+                        );
+                        const y = Math.max(
+                          0,
+                          Math.min(pos.y, imageSize.height - size),
+                        );
+
+                        return { x, y };
+                      }}
+                      onDragEnd={handleRectChange}
+                      onTransformEnd={handleRectChange}
+                    />
+                    <Transformer
+                      ref={transformerRef}
+                      rotateEnabled={false}
+                      enabledAnchors={[
+                        "top-left",
+                        "top-right",
+                        "bottom-left",
+                        "bottom-right",
+                      ]}
+                      boundBoxFunc={(_oldBox, newBox) => {
+                        const minSize = 50;
+                        const maxSize = Math.min(
+                          imageSize.width,
+                          imageSize.height,
+                        );
+
+                        // Clamp dimensions to stage bounds first
+                        const clampedWidth = Math.max(
+                          minSize,
+                          Math.min(newBox.width, maxSize),
+                        );
+                        const clampedHeight = Math.max(
+                          minSize,
+                          Math.min(newBox.height, maxSize),
+                        );
+
+                        // Enforce square using average
+                        const size = (clampedWidth + clampedHeight) / 2;
+
+                        // Clamp position to keep square within bounds
+                        const x = Math.max(
+                          0,
+                          Math.min(newBox.x, imageSize.width - size),
+                        );
+                        const y = Math.max(
+                          0,
+                          Math.min(newBox.y, imageSize.height - size),
+                        );
+
+                        return {
+                          ...newBox,
+                          x,
+                          y,
+                          width: size,
+                          height: size,
+                        };
+                      }}
+                    />
+                  </Layer>
+                </Stage>
+              </div>
+            ) : (
+              <div className="flex items-center justify-center text-muted-foreground">
+                {t("wizard.step2.selectCameraPrompt")}
+              </div>
+            )}
+          </div>
+        </div>
+      </div>
+
+      <div className="flex flex-col gap-3 pt-3 sm:flex-row sm:justify-end sm:gap-4">
+        <Button type="button" onClick={onBack} className="sm:flex-1">
+          {t("button.back", { ns: "common" })}
+        </Button>
+        <Button
+          type="button"
+          onClick={handleContinue}
+          variant="select"
+          className="flex items-center justify-center gap-2 sm:flex-1"
+          disabled={!canContinue}
+        >
+          {t("button.continue", { ns: "common" })}
+        </Button>
+      </div>
+    </div>
+  );
+}

web/src/components/classification/wizard/Step3ChooseExamples.tsx

@@ -0,0 +1,444 @@
+import { Button } from "@/components/ui/button";
+import { useTranslation } from "react-i18next";
+import { useState, useEffect, useCallback, useMemo } from "react";
+import ActivityIndicator from "@/components/indicators/activity-indicator";
+import axios from "axios";
+import { toast } from "sonner";
+import { Step1FormData } from "./Step1NameAndDefine";
+import { Step2FormData } from "./Step2StateArea";
+import useSWR from "swr";
+import { baseUrl } from "@/api/baseUrl";
+import { isMobile } from "react-device-detect";
+import { cn } from "@/lib/utils";
+
+export type Step3FormData = {
+  examplesGenerated: boolean;
+  imageClassifications?: { [imageName: string]: string };
+};
+
+type Step3ChooseExamplesProps = {
+  step1Data: Step1FormData;
+  step2Data?: Step2FormData;
+  initialData?: Partial<Step3FormData>;
+  onClose: () => void;
+  onBack: () => void;
+};
+
+export default function Step3ChooseExamples({
+  step1Data,
+  step2Data,
+  initialData,
+  onClose,
+  onBack,
+}: Step3ChooseExamplesProps) {
+  const { t } = useTranslation(["views/classificationModel"]);
+  const [isGenerating, setIsGenerating] = useState(false);
+  const [hasGenerated, setHasGenerated] = useState(
+    initialData?.examplesGenerated || false,
+  );
+  const [imageClassifications, setImageClassifications] = useState<{
+    [imageName: string]: string;
+  }>(initialData?.imageClassifications || {});
+  const [isTraining, setIsTraining] = useState(false);
+  const [isProcessing, setIsProcessing] = useState(false);
+  const [currentClassIndex, setCurrentClassIndex] = useState(0);
+  const [selectedImages, setSelectedImages] = useState<Set<string>>(new Set());
+
+  const { data: trainImages, mutate: refreshTrainImages } = useSWR<string[]>(
+    hasGenerated ? `classification/${step1Data.modelName}/train` : null,
+  );
+
+  const unknownImages = useMemo(() => {
+    if (!trainImages) return [];
+    return trainImages;
+  }, [trainImages]);
+
+  const toggleImageSelection = useCallback((imageName: string) => {
+    setSelectedImages((prev) => {
+      const newSet = new Set(prev);
+      if (newSet.has(imageName)) {
+        newSet.delete(imageName);
+      } else {
+        newSet.add(imageName);
+      }
+      return newSet;
+    });
+  }, []);
+
+  // Get all classes (excluding "none" - it will be auto-assigned)
+  const allClasses = useMemo(() => {
+    return [...step1Data.classes];
+  }, [step1Data.classes]);
+
+  const currentClass = allClasses[currentClassIndex];
+
+  const processClassificationsAndTrain = useCallback(
+    async (classifications: { [imageName: string]: string }) => {
+      // Step 1: Create config for the new model
+      const modelConfig: {
+        enabled: boolean;
+        name: string;
+        threshold: number;
+        state_config?: {
+          cameras: Record<string, { crop: number[] }>;
+          motion: boolean;
+        };
+        object_config?: { objects: string[]; classification_type: string };
+      } = {
+        enabled: true,
+        name: step1Data.modelName,
+        threshold: 0.8,
+      };
+
+      if (step1Data.modelType === "state") {
+        // State model config
+        const cameras: Record<string, { crop: number[] }> = {};
+        step2Data?.cameraAreas.forEach((area) => {
+          cameras[area.camera] = {
+            crop: area.crop,
+          };
+        });
+
+        modelConfig.state_config = {
+          cameras,
+          motion: true,
+        };
+      } else {
+        // Object model config
+        modelConfig.object_config = {
+          objects: step1Data.objectLabel ? [step1Data.objectLabel] : [],
+          classification_type: step1Data.objectType || "sub_label",
+        } as { objects: string[]; classification_type: string };
+      }
+
+      // Update config via config API
+      await axios.put("/config/set", {
+        requires_restart: 0,
+        update_topic: `config/classification/custom/${step1Data.modelName}`,
+        config_data: {
+          classification: {
+            custom: {
+              [step1Data.modelName]: modelConfig,
+            },
+          },
+        },
+      });
+
+      // Step 2: Classify each image by moving it to the correct category folder
+      const categorizePromises = Object.entries(classifications).map(
+        ([imageName, className]) => {
+          if (!className) return Promise.resolve();
+          return axios.post(
+            `/classification/${step1Data.modelName}/dataset/categorize`,
+            {
+              training_file: imageName,
+              category: className === "none" ? "none" : className,
+            },
+          );
+        },
+      );
+      await Promise.all(categorizePromises);
+
+      // Step 3: Kick off training
+      await axios.post(`/classification/${step1Data.modelName}/train`);
+
+      toast.success(t("wizard.step3.trainingStarted"));
+      setIsTraining(true);
+    },
+    [step1Data, step2Data, t],
+  );
+
+  const handleContinueClassification = useCallback(async () => {
+    // Mark selected images with current class
+    const newClassifications = { ...imageClassifications };
+    selectedImages.forEach((imageName) => {
+      newClassifications[imageName] = currentClass;
+    });
+
+    // Check if we're on the last class to select
+    const isLastClass = currentClassIndex === allClasses.length - 1;
+
+    if (isLastClass) {
+      // Assign remaining unclassified images
+      unknownImages.slice(0, 24).forEach((imageName) => {
+        if (!newClassifications[imageName]) {
+          // For state models with 2 classes, assign to the last class
+          // For object models, assign to "none"
+          if (step1Data.modelType === "state" && allClasses.length === 2) {
+            newClassifications[imageName] = allClasses[allClasses.length - 1];
+          } else {
+            newClassifications[imageName] = "none";
+          }
+        }
+      });
+
+      // All done, trigger training immediately
+      setImageClassifications(newClassifications);
+      setIsProcessing(true);
+
+      try {
+        await processClassificationsAndTrain(newClassifications);
+      } catch (error) {
+        const axiosError = error as {
+          response?: { data?: { message?: string; detail?: string } };
+          message?: string;
+        };
+        const errorMessage =
+          axiosError.response?.data?.message ||
+          axiosError.response?.data?.detail ||
+          axiosError.message ||
+          "Failed to classify images";
+
+        toast.error(
+          t("wizard.step3.errors.classifyFailed", { error: errorMessage }),
+        );
+        setIsProcessing(false);
+      }
+    } else {
+      // Move to next class
+      setImageClassifications(newClassifications);
+      setCurrentClassIndex((prev) => prev + 1);
+      setSelectedImages(new Set());
+    }
+  }, [
+    selectedImages,
+    currentClass,
+    currentClassIndex,
+    allClasses,
+    imageClassifications,
+    unknownImages,
+    step1Data,
+    processClassificationsAndTrain,
+    t,
+  ]);
+
+  const generateExamples = useCallback(async () => {
+    setIsGenerating(true);
+
+    try {
+      if (step1Data.modelType === "state") {
+        // For state models, use cameras and crop areas
+        if (!step2Data?.cameraAreas || step2Data.cameraAreas.length === 0) {
+          toast.error(t("wizard.step3.errors.noCameras"));
+          setIsGenerating(false);
+          return;
+        }
+
+        const cameras: { [key: string]: [number, number, number, number] } = {};
+        step2Data.cameraAreas.forEach((area) => {
+          cameras[area.camera] = area.crop;
+        });
+
+        await axios.post("/classification/generate_examples/state", {
+          model_name: step1Data.modelName,
+          cameras,
+        });
+      } else {
+        // For object models, use label
+        if (!step1Data.objectLabel) {
+          toast.error(t("wizard.step3.errors.noObjectLabel"));
+          setIsGenerating(false);
+          return;
+        }
+
+        // For now, use all enabled cameras
+        // TODO: In the future, we might want to let users select specific cameras
+        await axios.post("/classification/generate_examples/object", {
+          model_name: step1Data.modelName,
+          label: step1Data.objectLabel,
+        });
+      }
+
+      setHasGenerated(true);
+      toast.success(t("wizard.step3.generateSuccess"));
+
+      await refreshTrainImages();
+    } catch (error) {
+      const axiosError = error as {
+        response?: { data?: { message?: string; detail?: string } };
+        message?: string;
+      };
+      const errorMessage =
+        axiosError.response?.data?.message ||
+        axiosError.response?.data?.detail ||
+        axiosError.message ||
+        "Failed to generate examples";
+
+      toast.error(
+        t("wizard.step3.errors.generateFailed", { error: errorMessage }),
+      );
+    } finally {
+      setIsGenerating(false);
+    }
+  }, [step1Data, step2Data, t, refreshTrainImages]);
+
+  useEffect(() => {
+    if (!hasGenerated && !isGenerating) {
+      generateExamples();
+    }
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, []);
+
+  const handleContinue = useCallback(async () => {
+    setIsProcessing(true);
+    try {
+      await processClassificationsAndTrain(imageClassifications);
+    } catch (error) {
+      const axiosError = error as {
+        response?: { data?: { message?: string; detail?: string } };
+        message?: string;
+      };
+      const errorMessage =
+        axiosError.response?.data?.message ||
+        axiosError.response?.data?.detail ||
+        axiosError.message ||
+        "Failed to classify images";
+
+      toast.error(
+        t("wizard.step3.errors.classifyFailed", { error: errorMessage }),
+      );
+      setIsProcessing(false);
+    }
+  }, [imageClassifications, processClassificationsAndTrain, t]);
+
+  const unclassifiedImages = useMemo(() => {
+    if (!unknownImages) return [];
+    const images = unknownImages.slice(0, 24);
+
+    // Only filter if we have any classifications
+    if (Object.keys(imageClassifications).length === 0) {
+      return images;
+    }
+
+    return images.filter((img) => !imageClassifications[img]);
+  }, [unknownImages, imageClassifications]);
+
+  const allImagesClassified = useMemo(() => {
+    return unclassifiedImages.length === 0;
+  }, [unclassifiedImages]);
+
+  return (
+    <div className="flex flex-col gap-6">
+      {isTraining ? (
+        <div className="flex flex-col items-center gap-6 py-12">
+          <ActivityIndicator className="size-12" />
+          <div className="text-center">
+            <h3 className="mb-2 text-lg font-medium">
+              {t("wizard.step3.training.title")}
+            </h3>
+            <p className="text-sm text-muted-foreground">
+              {t("wizard.step3.training.description")}
+            </p>
+          </div>
+          <Button onClick={onClose} variant="select" className="mt-4">
+            {t("button.close", { ns: "common" })}
+          </Button>
+        </div>
+      ) : isGenerating ? (
+        <div className="flex h-[50vh] flex-col items-center justify-center gap-4">
+          <ActivityIndicator className="size-12" />
+          <div className="text-center">
+            <h3 className="mb-2 text-lg font-medium">
+              {t("wizard.step3.generating.title")}
+            </h3>
+            <p className="text-sm text-muted-foreground">
+              {t("wizard.step3.generating.description")}
+            </p>
+          </div>
+        </div>
+      ) : hasGenerated ? (
+        <div className="flex flex-col gap-4">
+          {!allImagesClassified && (
+            <div className="text-center">
+              <h3 className="text-lg font-medium">
+                {t("wizard.step3.selectImagesPrompt", {
+                  className: currentClass,
+                })}
+              </h3>
+              <p className="text-sm text-muted-foreground">
+                {t("wizard.step3.selectImagesDescription")}
+              </p>
+            </div>
+          )}
+          <div
+            className={cn(
+              "rounded-lg bg-secondary/30 p-4",
+              isMobile && "max-h-[60vh] overflow-y-auto",
+            )}
+          >
+            {!unknownImages || unknownImages.length === 0 ? (
+              <div className="flex h-[40vh] flex-col items-center justify-center gap-4">
+                <p className="text-muted-foreground">
+                  {t("wizard.step3.noImages")}
+                </p>
+                <Button onClick={generateExamples} variant="select">
+                  {t("wizard.step3.retryGenerate")}
+                </Button>
+              </div>
+            ) : allImagesClassified && isProcessing ? (
+              <div className="flex h-[40vh] flex-col items-center justify-center gap-4">
+                <ActivityIndicator className="size-12" />
+                <p className="text-lg font-medium">
+                  {t("wizard.step3.classifying")}
+                </p>
+              </div>
+            ) : (
+              <div className="grid grid-cols-2 gap-4 sm:grid-cols-6">
+                {unclassifiedImages.map((imageName, index) => {
+                  const isSelected = selectedImages.has(imageName);
+                  return (
+                    <div
+                      key={imageName}
+                      className={cn(
+                        "aspect-square cursor-pointer overflow-hidden rounded-lg border-2 bg-background transition-all",
+                        isSelected && "border-selected ring-2 ring-selected",
+                      )}
+                      onClick={() => toggleImageSelection(imageName)}
+                    >
+                      <img
+                        src={`${baseUrl}clips/${step1Data.modelName}/train/${imageName}`}
+                        alt={`Example ${index + 1}`}
+                        className="h-full w-full object-cover"
+                      />
+                    </div>
+                  );
+                })}
+              </div>
+            )}
+          </div>
+        </div>
+      ) : (
+        <div className="flex h-[50vh] flex-col items-center justify-center gap-4">
+          <p className="text-sm text-destructive">
+            {t("wizard.step3.errors.generationFailed")}
+          </p>
+          <Button onClick={generateExamples} variant="select">
+            {t("wizard.step3.retryGenerate")}
+          </Button>
+        </div>
+      )}
+
+      {!isTraining && (
+        <div className="flex flex-col gap-3 pt-3 sm:flex-row sm:justify-end sm:gap-4">
+          <Button type="button" onClick={onBack} className="sm:flex-1">
+            {t("button.back", { ns: "common" })}
+          </Button>
+          <Button
+            type="button"
+            onClick={
+              allImagesClassified
+                ? handleContinue
+                : handleContinueClassification
+            }
+            variant="select"
+            className="flex items-center justify-center gap-2 sm:flex-1"
+            disabled={!hasGenerated || isGenerating || isProcessing}
+          >
+            {isProcessing && <ActivityIndicator className="size-4" />}
+            {t("button.continue", { ns: "common" })}
+          </Button>
+        </div>
+      )}
+    </div>
+  );
+}

web/src/views/classification/ModelSelectionView.tsx

@@ -10,11 +10,14 @@ import {
   CustomClassificationModelConfig,
   FrigateConfig,
 } from "@/types/frigateConfig";
-import { useMemo, useState } from "react";
+import { useEffect, useMemo, useState } from "react";
 import { isMobile } from "react-device-detect";
 import { useTranslation } from "react-i18next";
 import { FaFolderPlus } from "react-icons/fa";
+import { MdModelTraining } from "react-icons/md";
 import useSWR from "swr";
+import Heading from "@/components/ui/heading";
+import { useOverlayState } from "@/hooks/use-overlay-state";
 
 const allModelTypes = ["objects", "states"] as const;
 type ModelType = (typeof allModelTypes)[number];
@@ -26,11 +29,24 @@ export default function ModelSelectionView({
   onClick,
 }: ModelSelectionViewProps) {
   const { t } = useTranslation(["views/classificationModel"]);
-  const [page, setPage] = useState<ModelType>("objects");
+  const [page, setPage] = useOverlayState<ModelType>("objects", "objects");
-  const [pageToggle, setPageToggle] = useOptimisticState(page, setPage, 100);
+  const [pageToggle, setPageToggle] = useOptimisticState(
-  const { data: config } = useSWR<FrigateConfig>("config", {
+    page || "objects",
+    setPage,
+    100,
+  );
+  const { data: config, mutate: refreshConfig } = useSWR<FrigateConfig>(
+    "config",
+    {
       revalidateOnFocus: false,
-  });
+    },
+  );
+
+  // title
+
+  useEffect(() => {
+    document.title = t("documentTitle");
+  }, [t]);
 
   // data
 
@@ -64,15 +80,15 @@ export default function ModelSelectionView({
     return <ActivityIndicator />;
   }
 
-  if (classificationConfigs.length == 0) {
-    return <div>You need to setup a custom model configuration.</div>;
-  }
-
   return (
     <div className="flex size-full flex-col p-2">
       <ClassificationModelWizardDialog
         open={newModel}
-        onClose={() => setNewModel(false)}
+        defaultModelType={pageToggle === "objects" ? "object" : "state"}
+        onClose={() => {
+          setNewModel(false);
+          refreshConfig();
+        }}
       />
 
       <div className="flex h-12 w-full items-center justify-between">
@@ -84,7 +100,6 @@ export default function ModelSelectionView({
             value={pageToggle}
             onValueChange={(value: ModelType) => {
               if (value) {
-                // Restrict viewer navigation
                 setPageToggle(value);
               }
             }}
@@ -117,13 +132,46 @@ export default function ModelSelectionView({
         </div>
       </div>
       <div className="flex size-full gap-2 p-2">
-        {selectedClassificationConfigs.map((config) => (
+        {selectedClassificationConfigs.length === 0 ? (
+          <NoModelsView
+            onCreateModel={() => setNewModel(true)}
+            modelType={pageToggle}
+          />
+        ) : (
+          selectedClassificationConfigs.map((config) => (
             <ModelCard
               key={config.name}
               config={config}
               onClick={() => onClick(config)}
             />
-        ))}
+          ))
+        )}
+      </div>
+    </div>
+  );
+}
+
+function NoModelsView({
+  onCreateModel,
+  modelType,
+}: {
+  onCreateModel: () => void;
+  modelType: ModelType;
+}) {
+  const { t } = useTranslation(["views/classificationModel"]);
+  const typeKey = modelType === "objects" ? "object" : "state";
+
+  return (
+    <div className="flex size-full items-center justify-center">
+      <div className="flex flex-col items-center gap-2">
+        <MdModelTraining className="size-8" />
+        <Heading as="h4">{t(`noModels.${typeKey}.title`)}</Heading>
+        <div className="mb-3 text-center text-secondary-foreground">
+          {t(`noModels.${typeKey}.description`)}
+        </div>
+        <Button size="sm" variant="select" onClick={onCreateModel}>
+          {t(`noModels.${typeKey}.buttonText`)}
+        </Button>
       </div>
     </div>
   );
@@ -139,13 +187,17 @@ function ModelCard({ config, onClick }: ModelCardProps) {
   }>(`classification/${config.name}/dataset`, { revalidateOnFocus: false });
 
   const coverImage = useMemo(() => {
-    if (!dataset?.length) {
+    if (!dataset) {
       return undefined;
     }
 
     const keys = Object.keys(dataset).filter((key) => key != "none");
     const selectedKey = keys[0];
 
+    if (!dataset[selectedKey]) {
+      return undefined;
+    }
+
     return {
       name: selectedKey,
       img: dataset[selectedKey][0],

web/src/views/classification/ModelTrainingView.tsx

@@ -642,6 +642,7 @@ function DatasetGrid({
               filepath: `clips/${modelName}/dataset/${categoryName}/${image}`,
               name: "",
             }}
+            showArea={false}
             selected={selectedImages.includes(image)}
             i18nLibrary="views/classificationModel"
             onClick={(data, _) => onClickImages([data.filename], true)}