Merge branch 'dev' into 230615-feature-camera-autoconf

.devcontainer/devcontainer.json

@@ -73,7 +73,11 @@
         "isort.args": ["--settings-path=./pyproject.toml"],
         "[python]": {
           "editor.defaultFormatter": "ms-python.black-formatter",
-          "editor.formatOnSave": true
+          "editor.formatOnSave": true,
+          "editor.codeActionsOnSave": {
+            "source.fixAll": true,
+            "source.organizeImports": true
+          }
         },
         "[json][jsonc]": {
           "editor.defaultFormatter": "esbenp.prettier-vscode"
@@ -86,7 +90,7 @@
           "editor.tabSize": 2
         },
         "cSpell.ignoreWords": ["rtmp"],
-        "cSpell.words": ["preact"]
+        "cSpell.words": ["preact", "astype", "hwaccel", "mqtt"]
       }
     }
   }

docs/docs/configuration/advanced.md

@@ -96,6 +96,16 @@ model:
 
 Note that if you rename objects in the labelmap, you will also need to update your `objects -> track` list as well.
 
+:::caution
+
+Some labels have special handling and modifications can disable functionality.
+
+`person` objects are associated with `face` and `amazon`
+
+`car` objects are associated with `license_plate`, `ups`, `fedex`, `amazon`
+
+:::
+
 ## Custom ffmpeg build
 
 Included with Frigate is a build of ffmpeg that works for the vast majority of users. However, there exists some hardware setups which have incompatibilities with the included build. In this case, a docker volume mapping can be used to overwrite the included ffmpeg build with an ffmpeg build that works for your specific hardware setup.

docs/docs/configuration/index.md

@@ -203,10 +203,10 @@ detect:
   max_disappeared: 25
   # Optional: Configuration for stationary object tracking
   stationary:
-    # Optional: Frequency for confirming stationary objects (default: shown below)
+    # Optional: Frequency for confirming stationary objects (default: same as threshold)
-    # When set to 0, object detection will not confirm stationary objects until movement is detected.
+    # When set to 1, object detection will run to confirm the object still exists on every frame.
     # If set to 10, object detection will run to confirm the object still exists on every 10th frame.
-    interval: 0
+    interval: 50
     # Optional: Number of frames without a position change for an object to be considered stationary (default: 10x the frame rate or 10s)
     threshold: 50
     # Optional: Define a maximum number of frames for tracking a stationary object (default: not set, track forever)
@@ -222,6 +222,20 @@ detect:
       # Optional: Object specific values
       objects:
         person: 1000
+  # Optional: Milliseconds to offset detect annotations by (default: shown below).
+  # There can often be latency between a recording and the detect process,
+  # especially when using separate streams for detect and record.
+  # Use this setting to make the timeline bounding boxes more closely align
+  # with the recording. The value can be positive or negative.
+  # TIP: Imagine there is an event clip with a person walking from left to right. 
+  #      If the event timeline bounding box is consistently to the left of the person
+  #      then the value should be decreased. Similarly, if a person is walking from
+  #      left to right and the bounding box is consistently ahead of the person
+  #      then the value should be increased.
+  # TIP: This offset is dynamic so you can change the value and it will update existing
+  #      events, this makes it easy to tune.
+  # WARNING: Fast moving objects will likely not have the bounding box align.
+  annotation_offset: 0
 
 # Optional: Object configuration
 # NOTE: Can be overridden at the camera level

docs/docs/configuration/stationary_objects.md

@@ -1,6 +1,6 @@
 # Stationary Objects
 
-An object is considered stationary when it is being tracked and has been in a very similar position for a certain number of frames. This number is defined in the configuration under `detect -> stationary -> threshold`, and is 10x the frame rate (or 10 seconds) by default. Once an object is considered stationary, it will remain stationary until motion occurs near the object at which point object detection will start running again. If the object changes location, it will be considered active.
+An object is considered stationary when it is being tracked and has been in a very similar position for a certain number of frames. This number is defined in the configuration under `detect -> stationary -> threshold`, and is 10x the frame rate (or 10 seconds) by default. Once an object is considered stationary, it will remain stationary until motion occurs within the object at which point object detection will start running again. If the object changes location, it will be considered active.
 
 ## Why does it matter if an object is stationary?
 
@@ -13,11 +13,11 @@ The default config is:
 ```yaml
 detect:
   stationary:
-    interval: 0
+    interval: 50
     threshold: 50
 ```
 
-`interval` is defined as the frequency for running detection on stationary objects. This means that by default once an object is considered stationary, detection will not be run on it until motion is detected. With `interval > 0`, every nth frames detection will be run to make sure the object is still there.
+`interval` is defined as the frequency for running detection on stationary objects. This means that by default once an object is considered stationary, detection will not be run on it until motion is detected or until the interval (every 50th frame by default). With `interval >= 1`, every nth frames detection will be run to make sure the object is still there.
 
 NOTE: There is no way to disable stationary object tracking with this value.
 

docs/docs/integrations/mqtt.md

@@ -62,7 +62,9 @@ Message published for each changed event. The first message is published when th
     "has_clip": false,
     "stationary": false, // whether or not the object is considered stationary
     "motionless_count": 0, // number of frames the object has been motionless
-    "position_changes": 2 // number of times the object has moved from a stationary position
+    "position_changes": 2, // number of times the object has moved from a stationary position
+    "attributes": [], // set of unique attributes that have been identified on the object
+    "current_attributes": [] // detailed data about the current attributes in this frame
   },
   "after": {
     "id": "1607123955.475377-mxklsc",
@@ -87,7 +89,16 @@ Message published for each changed event. The first message is published when th
     "has_clip": false,
     "stationary": false, // whether or not the object is considered stationary
     "motionless_count": 0, // number of frames the object has been motionless
-    "position_changes": 2 // number of times the object has changed position
+    "position_changes": 2, // number of times the object has changed position
+    "attributes": ["face"], // set of unique attributes that have been identified on the object
+    "current_attributes": [
+      // detailed data about the current attributes in this frame
+      {
+        "label": "face",
+        "box": [442, 506, 534, 524],
+        "score": 0.64
+      }
+    ]
   }
 }
 ```
@@ -163,9 +174,9 @@ Topic with current motion contour area for a camera. Published value is an integ
 
 Topic to send PTZ commands to camera.
 
-| Command                | Description                                                                             |
+| Command                | Description                                                                               |
-| ---------------------- | --------------------------------------------------------------------------------------- |
+| ---------------------- | ----------------------------------------------------------------------------------------- |
 | `preset-<preset_name>` | send command to move to preset with name `<preset_name>`                                  |
 | `MOVE_<dir>`           | send command to continuously move in `<dir>`, possible values are [UP, DOWN, LEFT, RIGHT] |
 | `ZOOM_<dir>`           | send command to continuously zoom `<dir>`, possible values are [IN, OUT]                  |
-| `STOP`                 | send command to stop moving                                                             |
+| `STOP`                 | send command to stop moving                                                               |

frigate/config.py

@@ -13,8 +13,6 @@ from pydantic.fields import PrivateAttr
 
 from frigate.const import CACHE_DIR, DEFAULT_DB_PATH, REGEX_CAMERA_NAME, YAML_EXT
 from frigate.detectors import DetectorConfig, ModelConfig
-from frigate.detectors.detector_config import InputTensorEnum  # noqa: F401
-from frigate.detectors.detector_config import PixelFormatEnum  # noqa: F401
 from frigate.detectors.detector_config import BaseDetectorConfig
 from frigate.ffmpeg_presets import (
     parse_preset_hardware_acceleration_decode,
@@ -191,7 +189,7 @@ class RecordConfig(FrigateBaseModel):
 
 class MotionConfig(FrigateBaseModel):
     threshold: int = Field(
-        default=40,
+        default=30,
         title="Motion detection threshold (1-255).",
         ge=1,
         le=255,
@@ -200,10 +198,10 @@ class MotionConfig(FrigateBaseModel):
         default=0.8, title="Lightning detection threshold (0.3-1.0).", ge=0.3, le=1.0
     )
     improve_contrast: bool = Field(default=True, title="Improve Contrast")
-    contour_area: Optional[int] = Field(default=15, title="Contour Area")
+    contour_area: Optional[int] = Field(default=10, title="Contour Area")
     delta_alpha: float = Field(default=0.2, title="Delta Alpha")
     frame_alpha: float = Field(default=0.02, title="Frame Alpha")
-    frame_height: Optional[int] = Field(default=50, title="Frame Height")
+    frame_height: Optional[int] = Field(default=100, title="Frame Height")
     mask: Union[str, List[str]] = Field(
         default="", title="Coordinates polygon for the motion mask."
     )
@@ -253,9 +251,8 @@ class StationaryMaxFramesConfig(FrigateBaseModel):
 
 class StationaryConfig(FrigateBaseModel):
     interval: Optional[int] = Field(
-        default=0,
         title="Frame interval for checking stationary objects.",
-        ge=0,
+        gt=0,
     )
     threshold: Optional[int] = Field(
         title="Number of frames without a position change for an object to be considered stationary",
@@ -988,6 +985,9 @@ class FrigateConfig(FrigateBaseModel):
             stationary_threshold = camera_config.detect.fps * 10
             if camera_config.detect.stationary.threshold is None:
                 camera_config.detect.stationary.threshold = stationary_threshold
+            # default to the stationary_threshold if not defined
+            if camera_config.detect.stationary.interval is None:
+                camera_config.detect.stationary.interval = stationary_threshold
 
             # FFMPEG input substitution
             for input in camera_config.ffmpeg.inputs:

frigate/events/maintainer.py

@@ -147,6 +147,23 @@ class EventProcessor(threading.Thread):
                 )
             )
 
+            attributes = [
+                (
+                    None
+                    if event_data["snapshot"] is None
+                    else {
+                        "box": to_relative_box(
+                            width,
+                            height,
+                            a["box"],
+                        ),
+                        "label": a["label"],
+                        "score": a["score"],
+                    }
+                )
+                for a in event_data["snapshot"]["attributes"]
+            ]
+
             # keep these from being set back to false because the event
             # may have started while recordings and snapshots were enabled
             # this would be an issue for long running events
@@ -173,9 +190,14 @@ class EventProcessor(threading.Thread):
                     "region": region,
                     "score": score,
                     "top_score": event_data["top_score"],
+                    "attributes": attributes,
                 },
             }
 
+            # only overwrite the sub_label in the database if it's set
+            if event_data.get("sub_label") is not None:
+                event[Event.sub_label] = event_data["sub_label"]
+
             (
                 Event.insert(event)
                 .on_conflict(

frigate/motion/improved_motion.py

@@ -38,6 +38,7 @@ class ImprovedMotionDetector(MotionDetector):
         self.improve_contrast = improve_contrast
         self.threshold = threshold
         self.contour_area = contour_area
+        self.clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
 
     def detect(self, frame):
         motion_boxes = []
@@ -55,7 +56,7 @@ class ImprovedMotionDetector(MotionDetector):
 
         # Improve contrast
         if self.improve_contrast.value:
-            resized_frame = cv2.equalizeHist(resized_frame)
+            resized_frame = self.clahe.apply(resized_frame)
 
         # mask frame
         resized_frame[self.mask] = [255]

frigate/object_detection.py

@@ -10,8 +10,8 @@ from abc import ABC, abstractmethod
 import numpy as np
 from setproctitle import setproctitle
 
-from frigate.config import InputTensorEnum
 from frigate.detectors import create_detector
+from frigate.detectors.detector_config import InputTensorEnum
 from frigate.util import EventsPerSecond, SharedMemoryFrameManager, listen, load_labels
 
 logger = logging.getLogger(__name__)

frigate/object_processing.py

@@ -24,6 +24,7 @@ from frigate.const import CLIPS_DIR
 from frigate.events.maintainer import EventTypeEnum
 from frigate.util import (
     SharedMemoryFrameManager,
+    area,
     calculate_region,
     draw_box_with_label,
     draw_timestamp,
@@ -42,11 +43,45 @@ def on_edge(box, frame_shape):
         return True
 
 
-def is_better_thumbnail(current_thumb, new_obj, frame_shape) -> bool:
+def has_better_attr(current_thumb, new_obj, attr_label) -> bool:
+    max_new_attr = max(
+        [0]
+        + [area(a["box"]) for a in new_obj["attributes"] if a["label"] == attr_label]
+    )
+    max_current_attr = max(
+        [0]
+        + [
+            area(a["box"])
+            for a in current_thumb["attributes"]
+            if a["label"] == attr_label
+        ]
+    )
+
+    # if the thumb has a higher scoring attr
+    return max_new_attr > max_current_attr
+
+
+def is_better_thumbnail(label, current_thumb, new_obj, frame_shape) -> bool:
     # larger is better
     # cutoff images are less ideal, but they should also be smaller?
     # better scores are obviously better too
 
+    # check face on person
+    if label == "person":
+        if has_better_attr(current_thumb, new_obj, "face"):
+            return True
+        # if the current thumb has a face attr, dont update unless it gets better
+        if any([a["label"] == "face" for a in current_thumb["attributes"]]):
+            return False
+
+    # check license_plate on car
+    if label == "car":
+        if has_better_attr(current_thumb, new_obj, "license_plate"):
+            return True
+        # if the current thumb has a license_plate attr, dont update unless it gets better
+        if any([a["label"] == "license_plate" for a in current_thumb["attributes"]]):
+            return False
+
     # if the new_thumb is on an edge, and the current thumb is not
     if on_edge(new_obj["box"], frame_shape) and not on_edge(
         current_thumb["box"], frame_shape
@@ -76,6 +111,7 @@ class TrackedObject:
         self.zone_presence = {}
         self.current_zones = []
         self.entered_zones = []
+        self.attributes = set()
         self.false_positive = True
         self.has_clip = False
         self.has_snapshot = False
@@ -125,7 +161,10 @@ class TrackedObject:
         if not self.false_positive:
             # determine if this frame is a better thumbnail
             if self.thumbnail_data is None or is_better_thumbnail(
-                self.thumbnail_data, obj_data, self.camera_config.frame_shape
+                self.obj_data["label"],
+                self.thumbnail_data,
+                obj_data,
+                self.camera_config.frame_shape,
             ):
                 self.thumbnail_data = {
                     "frame_time": obj_data["frame_time"],
@@ -133,6 +172,7 @@ class TrackedObject:
                     "area": obj_data["area"],
                     "region": obj_data["region"],
                     "score": obj_data["score"],
+                    "attributes": obj_data["attributes"],
                 }
                 thumb_update = True
 
@@ -164,6 +204,19 @@ class TrackedObject:
                 if 0 < zone_score < zone.inertia:
                     self.zone_presence[name] = zone_score - 1
 
+        # maintain attributes
+        for attr in obj_data["attributes"]:
+            self.attributes.add(attr["label"])
+
+        # populate the sub_label for car with first logo if it exists
+        if self.obj_data["label"] == "car" and "sub_label" not in self.obj_data:
+            recognized_logos = self.attributes.intersection(
+                set(["ups", "fedex", "amazon"])
+            )
+            if len(recognized_logos) > 0:
+                self.obj_data["sub_label"] = recognized_logos.pop()
+
+        # check for significant change
         if not self.false_positive:
             # if the zones changed, signal an update
             if set(self.current_zones) != set(current_zones):
@@ -214,6 +267,8 @@ class TrackedObject:
             "entered_zones": self.entered_zones.copy(),
             "has_clip": self.has_clip,
             "has_snapshot": self.has_snapshot,
+            "attributes": list(self.attributes),
+            "current_attributes": self.obj_data["attributes"],
         }
 
         if include_thumbnail:
@@ -294,6 +349,21 @@ class TrackedObject:
                 color=color,
             )
 
+            # draw any attributes
+            for attribute in self.thumbnail_data["attributes"]:
+                box = attribute["box"]
+                draw_box_with_label(
+                    best_frame,
+                    box[0],
+                    box[1],
+                    box[2],
+                    box[3],
+                    attribute["label"],
+                    f"{attribute['score']:.0%}",
+                    thickness=thickness,
+                    color=color,
+                )
+
         if crop:
             box = self.thumbnail_data["box"]
             box_size = 300
@@ -421,6 +491,21 @@ class CameraState:
                     color=color,
                 )
 
+                # draw any attributes
+                for attribute in obj["current_attributes"]:
+                    box = attribute["box"]
+                    draw_box_with_label(
+                        frame_copy,
+                        box[0],
+                        box[1],
+                        box[2],
+                        box[3],
+                        attribute["label"],
+                        f"{attribute['score']:.0%}",
+                        thickness=thickness,
+                        color=color,
+                    )
+
         if draw_options.get("regions"):
             for region in regions:
                 cv2.rectangle(
@@ -553,6 +638,7 @@ class CameraState:
                 # or the current object is older than desired, use the new object
                 if (
                     is_better_thumbnail(
+                        object_type,
                         current_best.thumbnail_data,
                         obj.thumbnail_data,
                         self.camera_config.frame_shape,

frigate/output.py

@@ -1,6 +1,7 @@
 import datetime
 import glob
 import logging
+import math
 import multiprocessing as mp
 import os
 import queue
@@ -269,161 +270,118 @@ class BirdsEyeFrameManager:
     def update_frame(self):
         """Update to a new frame for birdseye."""
 
-        def calculate_two_cam_layout(canvas, cameras_to_add: list[str]) -> tuple[any]:
-            """Calculate the optimal layout for 2 cameras."""
-            first_camera = cameras_to_add[0]
-            first_camera_dims = self.cameras[first_camera]["dimensions"].copy()
-            second_camera = cameras_to_add[1]
-            second_camera_dims = self.cameras[second_camera]["dimensions"].copy()
-
-            # check for optimal layout
-            if first_camera_dims[0] + second_camera_dims[0] < canvas_width:
-                # place cameras horizontally
-                first_scaled_width = int(
-                    canvas_height * first_camera_dims[0] / first_camera_dims[1]
-                )
-                second_scaled_width = int(
-                    canvas_height * second_camera_dims[0] / second_camera_dims[1]
-                )
-                first_height = canvas_height
-                second_height = canvas_height
-
-                if first_scaled_width + second_scaled_width > canvas_width:
-                    if first_scaled_width > second_scaled_width:
-                        first_scaled_width = canvas_width - second_scaled_width
-                        first_height = int(
-                            first_scaled_width
-                            * first_camera_dims[1]
-                            / first_camera_dims[0]
-                        )
-                    else:
-                        second_scaled_width = canvas_width - first_scaled_width
-                        second_height = int(
-                            second_scaled_width
-                            * second_camera_dims[1]
-                            / second_camera_dims[0]
-                        )
-
-                return [
-                    [
-                        (
-                            first_camera,
-                            (0, 0, first_scaled_width, first_height),
-                        ),
-                        (
-                            second_camera,
-                            (
-                                first_scaled_width + 1,
-                                0,
-                                second_scaled_width,
-                                second_height,
-                            ),
-                        ),
-                    ],
-                ]
-            else:
-                # place cameras vertically
-                top_scaled_width = int(
-                    (canvas_height / 2) * first_camera_dims[0] / first_camera_dims[1]
-                )
-                bottom_scaled_width = int(
-                    (canvas_height / 2) * second_camera_dims[0] / second_camera_dims[1]
-                )
-                return [
-                    [
-                        (
-                            first_camera,
-                            (0, 0, top_scaled_width, int(canvas_height / 2)),
-                        )
-                    ],
-                    [
-                        (
-                            second_camera,
-                            (
-                                0,
-                                int(canvas_height / 2),
-                                bottom_scaled_width,
-                                int(canvas_height / 2),
-                            ),
-                        )
-                    ],
-                ]
-
         def calculate_layout(
             canvas, cameras_to_add: list[str], coefficient
         ) -> tuple[any]:
-            """Calculate the optimal layout for 3+ cameras."""
+            """Calculate the optimal layout for 2+ cameras."""
             camera_layout: list[list[any]] = []
             camera_layout.append([])
-            canvas_aspect = canvas[0] / canvas[1]
+            canvas_gcd = math.gcd(canvas[0], canvas[1])
+            canvas_aspect_x = (canvas[0] / canvas_gcd) * coefficient
+            canvas_aspect_y = (canvas[0] / canvas_gcd) * coefficient
             starting_x = 0
             x = starting_x
             y = 0
             y_i = 0
-            max_height = 0
+            max_y = 0
             for camera in cameras_to_add:
                 camera_dims = self.cameras[camera]["dimensions"].copy()
-                camera_aspect = camera_dims[0] / camera_dims[1]
+                camera_gcd = math.gcd(camera_dims[0], camera_dims[1])
+                camera_aspect_x = camera_dims[0] / camera_gcd
+                camera_aspect_y = camera_dims[1] / camera_gcd
+
+                if round(camera_aspect_x / camera_aspect_y, 1) == 1.8:
+                    # account for slightly off 16:9 cameras
+                    camera_aspect_x = 16
+                    camera_aspect_y = 9
+                elif round(camera_aspect_x / camera_aspect_y, 1) == 1.3:
+                    # make 4:3 cameras the same relative size as 16:9
+                    camera_aspect_x = 12
+                    camera_aspect_y = 9
 
                 if camera_dims[1] > camera_dims[0]:
                     portrait = True
-                elif camera_aspect < canvas_aspect:
-                    # if the camera aspect ratio is less than canvas aspect ratio, it needs to be scaled down to fit
-                    camera_dims[0] *= camera_aspect / canvas_aspect
-                    camera_dims[1] *= camera_aspect / canvas_aspect
-                    portrait = False
                 else:
                     portrait = False
 
-                if (x + camera_dims[0] * coefficient) <= canvas[0]:
+                if (x + camera_aspect_x) <= canvas_aspect_x:
                     # insert if camera can fit on current row
-                    scaled_width = int(camera_dims[0] * coefficient)
                     camera_layout[y_i].append(
                         (
                             camera,
                             (
-                                x,
+                                camera_aspect_x,
-                                y,
+                                camera_aspect_y,
-                                scaled_width,
-                                int(camera_dims[1] * coefficient),
                             ),
                         )
                     )
-                    x += scaled_width
 
                     if portrait:
-                        starting_x = scaled_width
+                        starting_x = camera_aspect_x
                     else:
-                        max_height = max(
+                        max_y = max(
-                            max_height,
+                            max_y,
-                            int(camera_dims[1] * coefficient),
+                            camera_aspect_y,
                         )
+
+                    x += camera_aspect_x
                 else:
                     # move on to the next row and insert
-                    y += max_height
+                    y += max_y
                     y_i += 1
                     camera_layout.append([])
                     x = starting_x
 
-                    if camera_dims[0] * coefficient > canvas_width:
+                    if x + camera_aspect_x > canvas_aspect_x:
-                        safe_coefficient = 1
+                        return None
-                    else:
-                        safe_coefficient = coefficient
 
                     camera_layout[y_i].append(
                         (
                             camera,
-                            (
+                            (camera_aspect_x, camera_aspect_y),
-                                x,
-                                y,
-                                int(camera_dims[0] * safe_coefficient),
-                                int(camera_dims[1] * safe_coefficient),
-                            ),
                         )
                     )
-                    x += int(camera_dims[0] * safe_coefficient)
+                    x += camera_aspect_x
 
-            return (camera_layout, y + max_height)
+            if y + max_y > canvas_aspect_y:
+                return None
+
+            row_height = int(canvas_height / coefficient)
+
+            final_camera_layout = []
+            starting_x = 0
+            y = 0
+
+            for row in camera_layout:
+                final_row = []
+                x = starting_x
+                for cameras in row:
+                    camera_dims = self.cameras[cameras[0]]["dimensions"].copy()
+
+                    if camera_dims[1] > camera_dims[0]:
+                        scaled_height = int(row_height * coefficient)
+                        scaled_width = int(
+                            scaled_height * camera_dims[0] / camera_dims[1]
+                        )
+                        starting_x = scaled_width
+                    else:
+                        scaled_height = row_height
+                        scaled_width = int(
+                            scaled_height * camera_dims[0] / camera_dims[1]
+                        )
+
+                    if (
+                        x + scaled_width > canvas_width
+                        or y + scaled_height > canvas_height
+                    ):
+                        return None
+
+                    final_row.append((cameras[0], (x, y, scaled_width, scaled_height)))
+                    x += scaled_width
+                y += row_height
+                final_camera_layout.append(final_row)
+
+            return final_camera_layout
 
         # determine how many cameras are tracking objects within the last 30 seconds
         active_cameras = set(
@@ -493,30 +451,28 @@ class BirdsEyeFrameManager:
                         )
                     ]
                 ]
-            elif len(active_cameras) == 2:
-                self.camera_layout = calculate_two_cam_layout(
-                    (canvas_width, canvas_height), active_cameras_to_add
-                )
             else:
                 # calculate optimal layout
-                coefficient = 1.0
+                coefficient = 2
                 calculating = True
 
                 # decrease scaling coefficient until height of all cameras can fit into the birdseye canvas
                 while calculating:
-                    layout_candidate, total_height = calculate_layout(
+                    layout_candidate = calculate_layout(
                         (canvas_width, canvas_height),
                         active_cameras_to_add,
                         coefficient,
                     )
 
-                    if (canvas_height * 0.75) < total_height <= canvas_height:
+                    if not layout_candidate:
-                        calculating = False
+                        if coefficient < 10:
-                    elif total_height < canvas_height * 0.75:
+                            coefficient += 1
-                        coefficient += 0.1
+                            continue
-                        calculating = False
+                        else:
-                    else:
+                            logger.error("Error finding appropriate birdseye layout")
-                        coefficient -= 0.1
+                            return
+
+                    calculating = False
 
                 self.camera_layout = layout_candidate
 

frigate/record/maintainer.py

@@ -1,5 +1,6 @@
 """Maintain recording segments in cache."""
 
+import asyncio
 import datetime
 import logging
 import multiprocessing as mp
@@ -20,7 +21,7 @@ from frigate.config import FrigateConfig, RetainModeEnum
 from frigate.const import CACHE_DIR, MAX_SEGMENT_DURATION, RECORD_DIR
 from frigate.models import Event, Recordings
 from frigate.types import RecordMetricsTypes
-from frigate.util import area
+from frigate.util import area, get_video_properties
 
 logger = logging.getLogger(__name__)
 
@@ -42,7 +43,7 @@ class RecordingMaintainer(threading.Thread):
         self.recordings_info: dict[str, Any] = defaultdict(list)
         self.end_time_cache: dict[str, Tuple[datetime.datetime, float]] = {}
 
-    def move_files(self) -> None:
+    async def move_files(self) -> None:
         cache_files = sorted(
             [
                 d
@@ -121,115 +122,100 @@ class RecordingMaintainer(threading.Thread):
                 )
                 .order_by(Event.start_time)
             )
-            for r in recordings:
-                cache_path = r["cache_path"]
-                start_time = r["start_time"]
 
-                # Just delete files if recordings are turned off
+            await asyncio.gather(
-                if (
+                *(self.validate_and_move_segment(camera, events, r) for r in recordings)
-                    camera not in self.config.cameras
+            )
-                    or not self.process_info[camera]["record_enabled"].value
+
-                ):
+    async def validate_and_move_segment(
+        self, camera: str, events: Event, recording: dict[str, any]
+    ) -> None:
+        cache_path = recording["cache_path"]
+        start_time = recording["start_time"]
+
+        # Just delete files if recordings are turned off
+        if (
+            camera not in self.config.cameras
+            or not self.process_info[camera]["record_enabled"].value
+        ):
+            Path(cache_path).unlink(missing_ok=True)
+            self.end_time_cache.pop(cache_path, None)
+            return
+
+        if cache_path in self.end_time_cache:
+            end_time, duration = self.end_time_cache[cache_path]
+        else:
+            segment_info = get_video_properties(cache_path, get_duration=True)
+
+            if segment_info["duration"]:
+                duration = float(segment_info["duration"])
+            else:
+                duration = -1
+
+            # ensure duration is within expected length
+            if 0 < duration < MAX_SEGMENT_DURATION:
+                end_time = start_time + datetime.timedelta(seconds=duration)
+                self.end_time_cache[cache_path] = (end_time, duration)
+            else:
+                if duration == -1:
+                    logger.warning(f"Failed to probe corrupt segment {cache_path}")
+
+                logger.warning(f"Discarding a corrupt recording segment: {cache_path}")
+                Path(cache_path).unlink(missing_ok=True)
+                return
+
+        # if cached file's start_time is earlier than the retain days for the camera
+        if start_time <= (
+            (
+                datetime.datetime.now()
+                - datetime.timedelta(
+                    days=self.config.cameras[camera].record.retain.days
+                )
+            )
+        ):
+            # if the cached segment overlaps with the events:
+            overlaps = False
+            for event in events:
+                # if the event starts in the future, stop checking events
+                # and remove this segment
+                if event.start_time > end_time.timestamp():
+                    overlaps = False
                     Path(cache_path).unlink(missing_ok=True)
                     self.end_time_cache.pop(cache_path, None)
-                    continue
+                    break
 
-                if cache_path in self.end_time_cache:
+                # if the event is in progress or ends after the recording starts, keep it
-                    end_time, duration = self.end_time_cache[cache_path]
+                # and stop looking at events
-                else:
+                if event.end_time is None or event.end_time >= start_time.timestamp():
-                    ffprobe_cmd = [
+                    overlaps = True
-                        "ffprobe",
+                    break
-                        "-v",
-                        "error",
-                        "-show_entries",
-                        "format=duration",
-                        "-of",
-                        "default=noprint_wrappers=1:nokey=1",
-                        f"{cache_path}",
-                    ]
-                    p = sp.run(ffprobe_cmd, capture_output=True)
-                    if p.returncode == 0 and p.stdout.decode():
-                        duration = float(p.stdout.decode().strip())
-                    else:
-                        duration = -1
 
-                    # ensure duration is within expected length
+            if overlaps:
-                    if 0 < duration < MAX_SEGMENT_DURATION:
+                record_mode = self.config.cameras[camera].record.events.retain.mode
-                        end_time = start_time + datetime.timedelta(seconds=duration)
+                # move from cache to recordings immediately
-                        self.end_time_cache[cache_path] = (end_time, duration)
+                self.store_segment(
-                    else:
+                    camera,
-                        if duration == -1:
+                    start_time,
-                            logger.warning(
+                    end_time,
-                                f"Failed to probe corrupt segment {cache_path} : {p.returncode} - {str(p.stderr)}"
+                    duration,
-                            )
+                    cache_path,
-
+                    record_mode,
-                        logger.warning(
+                )
-                            f"Discarding a corrupt recording segment: {cache_path}"
+            # if it doesn't overlap with an event, go ahead and drop the segment
-                        )
+            # if it ends more than the configured pre_capture for the camera
-                        Path(cache_path).unlink(missing_ok=True)
+            else:
-                        continue
+                pre_capture = self.config.cameras[camera].record.events.pre_capture
-
+                most_recently_processed_frame_time = self.recordings_info[camera][-1][0]
-                # if cached file's start_time is earlier than the retain days for the camera
+                retain_cutoff = most_recently_processed_frame_time - pre_capture
-                if start_time <= (
+                if end_time.timestamp() < retain_cutoff:
-                    (
+                    Path(cache_path).unlink(missing_ok=True)
-                        datetime.datetime.now()
+                    self.end_time_cache.pop(cache_path, None)
-                        - datetime.timedelta(
+        # else retain days includes this segment
-                            days=self.config.cameras[camera].record.retain.days
+        else:
-                        )
+            record_mode = self.config.cameras[camera].record.retain.mode
-                    )
+            self.store_segment(
-                ):
+                camera, start_time, end_time, duration, cache_path, record_mode
-                    # if the cached segment overlaps with the events:
+            )
-                    overlaps = False
-                    for event in events:
-                        # if the event starts in the future, stop checking events
-                        # and remove this segment
-                        if event.start_time > end_time.timestamp():
-                            overlaps = False
-                            Path(cache_path).unlink(missing_ok=True)
-                            self.end_time_cache.pop(cache_path, None)
-                            break
-
-                        # if the event is in progress or ends after the recording starts, keep it
-                        # and stop looking at events
-                        if (
-                            event.end_time is None
-                            or event.end_time >= start_time.timestamp()
-                        ):
-                            overlaps = True
-                            break
-
-                    if overlaps:
-                        record_mode = self.config.cameras[
-                            camera
-                        ].record.events.retain.mode
-                        # move from cache to recordings immediately
-                        self.store_segment(
-                            camera,
-                            start_time,
-                            end_time,
-                            duration,
-                            cache_path,
-                            record_mode,
-                        )
-                    # if it doesn't overlap with an event, go ahead and drop the segment
-                    # if it ends more than the configured pre_capture for the camera
-                    else:
-                        pre_capture = self.config.cameras[
-                            camera
-                        ].record.events.pre_capture
-                        most_recently_processed_frame_time = self.recordings_info[
-                            camera
-                        ][-1][0]
-                        retain_cutoff = most_recently_processed_frame_time - pre_capture
-                        if end_time.timestamp() < retain_cutoff:
-                            Path(cache_path).unlink(missing_ok=True)
-                            self.end_time_cache.pop(cache_path, None)
-                # else retain days includes this segment
-                else:
-                    record_mode = self.config.cameras[camera].record.retain.mode
-                    self.store_segment(
-                        camera, start_time, end_time, duration, cache_path, record_mode
-                    )
 
     def segment_stats(
         self, camera: str, start_time: datetime.datetime, end_time: datetime.datetime
@@ -386,7 +372,7 @@ class RecordingMaintainer(threading.Thread):
                     break
 
             try:
-                self.move_files()
+                asyncio.run(self.move_files())
             except Exception as e:
                 logger.error(
                     "Error occurred when attempting to maintain recording cache"

frigate/test/test_config.py

@@ -730,7 +730,7 @@ class TestConfig(unittest.TestCase):
         assert config == frigate_config.dict(exclude_unset=True)
 
         runtime_config = frigate_config.runtime_config()
-        assert runtime_config.cameras["back"].motion.frame_height == 50
+        assert runtime_config.cameras["back"].motion.frame_height == 100
 
     def test_motion_contour_area_dynamic(self):
         config = {
@@ -758,7 +758,7 @@ class TestConfig(unittest.TestCase):
         assert config == frigate_config.dict(exclude_unset=True)
 
         runtime_config = frigate_config.runtime_config()
-        assert round(runtime_config.cameras["back"].motion.contour_area) == 15
+        assert round(runtime_config.cameras["back"].motion.contour_area) == 10
 
     def test_merge_labelmap(self):
         config = {

frigate/test/test_object_detector.py

@@ -6,8 +6,9 @@ from pydantic import parse_obj_as
 
 import frigate.detectors as detectors
 import frigate.object_detection
-from frigate.config import DetectorConfig, InputTensorEnum, ModelConfig
+from frigate.config import DetectorConfig, ModelConfig
 from frigate.detectors import DetectorTypeEnum
+from frigate.detectors.detector_config import InputTensorEnum
 
 
 class TestLocalObjectDetector(unittest.TestCase):

frigate/track/norfair_tracker.py

@@ -91,9 +91,13 @@ class NorfairTracker(ObjectTracker):
             "ymax": self.detect_config.height,
         }
 
-    def deregister(self, id):
+    def deregister(self, id, track_id):
         del self.tracked_objects[id]
         del self.disappeared[id]
+        self.tracker.tracked_objects = [
+            o for o in self.tracker.tracked_objects if o.global_id != track_id
+        ]
+        del self.track_id_map[track_id]
 
     # tracks the current position of the object based on the last N bounding boxes
     # returns False if the object has moved outside its previous position
@@ -167,7 +171,7 @@ class NorfairTracker(ObjectTracker):
         if self.update_position(id, obj["box"]):
             self.tracked_objects[id]["motionless_count"] += 1
             if self.is_expired(id):
-                self.deregister(id)
+                self.deregister(id, track_id)
                 return
         else:
             # register the first position change and then only increment if
@@ -261,8 +265,7 @@ class NorfairTracker(ObjectTracker):
         # clear expired tracks
         expired_ids = [k for k in self.track_id_map.keys() if k not in active_ids]
         for e_id in expired_ids:
-            self.deregister(self.track_id_map[e_id])
+            self.deregister(self.track_id_map[e_id], e_id)
-            del self.track_id_map[e_id]
 
     def debug_draw(self, frame, frame_time):
         active_detections = [

frigate/util.py

@@ -1147,31 +1147,66 @@ def to_relative_box(
 
 
 def get_video_properties(url, get_duration=False):
+    def calculate_duration(video: Optional[any]) -> float:
+        duration = None
+
+        if video is not None:
+            # Get the frames per second (fps) of the video stream
+            fps = video.get(cv2.CAP_PROP_FPS)
+            total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
+
+            if fps and total_frames:
+                duration = total_frames / fps
+
+        # if cv2 failed need to use ffprobe
+        if duration is None:
+            ffprobe_cmd = [
+                "ffprobe",
+                "-v",
+                "error",
+                "-show_entries",
+                "format=duration",
+                "-of",
+                "default=noprint_wrappers=1:nokey=1",
+                f"{url}",
+            ]
+            p = sp.run(ffprobe_cmd, capture_output=True)
+
+            if p.returncode == 0 and p.stdout.decode():
+                duration = float(p.stdout.decode().strip())
+            else:
+                duration = -1
+
+        return duration
+
     width = height = 0
-    # Open the video stream
-    video = cv2.VideoCapture(url)
 
-    # Check if the video stream was opened successfully
+    try:
-    if not video.isOpened():
+        # Open the video stream
-        logger.debug(f"Error opening video stream {url}.")
+        video = cv2.VideoCapture(url)
-        return None
 
-    # Get the width of frames in the video stream
+        # Check if the video stream was opened successfully
-    width = video.get(cv2.CAP_PROP_FRAME_WIDTH)
+        if not video.isOpened():
+            video = None
+    except Exception:
+        video = None
 
-    # Get the height of frames in the video stream
+    result = {}
-    height = video.get(cv2.CAP_PROP_FRAME_HEIGHT)
 
-    # Release the video stream
-    video.release()
-
-    result = {"width": round(width), "height": round(height)}
     if get_duration:
-        # Get the frames per second (fps) of the video stream
+        result["duration"] = calculate_duration(video)
-        fps = video.get(cv2.CAP_PROP_FPS)
-        total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
-        duration = total_frames / fps
 
-        result["duration"] = duration
+    if video is not None:
+        # Get the width of frames in the video stream
+        width = video.get(cv2.CAP_PROP_FRAME_WIDTH)
+
+        # Get the height of frames in the video stream
+        height = video.get(cv2.CAP_PROP_FRAME_HEIGHT)
+
+        # Release the video stream
+        video.release()
+
+        result["width"] = round(width)
+        result["height"] = round(height)
 
     return result

frigate/video.py

@@ -14,8 +14,9 @@ import cv2
 import numpy as np
 from setproctitle import setproctitle
 
-from frigate.config import CameraConfig, DetectConfig, PixelFormatEnum
+from frigate.config import CameraConfig, DetectConfig
 from frigate.const import CACHE_DIR
+from frigate.detectors.detector_config import PixelFormatEnum
 from frigate.log import LogPipe
 from frigate.motion import MotionDetector
 from frigate.motion.improved_motion import ImprovedMotionDetector
@@ -722,6 +723,14 @@ def process_frames(
     stop_event,
     exit_on_empty: bool = False,
 ):
+    # attribute labels are not tracked and are not assigned regions
+    attribute_label_map = {
+        "person": ["face", "amazon"],
+        "car": ["ups", "fedex", "amazon", "license_plate"],
+    }
+    all_attribute_labels = [
+        item for sublist in attribute_label_map.values() for item in sublist
+    ]
     fps = process_info["process_fps"]
     detection_fps = process_info["detection_fps"]
     current_frame_time = process_info["detection_frame"]
@@ -757,6 +766,7 @@ def process_frames(
         motion_boxes = motion_detector.detect(frame) if motion_enabled.value else []
 
         regions = []
+        consolidated_detections = []
 
         # if detection is disabled
         if not detection_enabled.value:
@@ -769,8 +779,8 @@ def process_frames(
             stationary_object_ids = [
                 obj["id"]
                 for obj in object_tracker.tracked_objects.values()
-                # if there hasn't been motion for 10 frames
+                # if it has exceeded the stationary threshold
-                if obj["motionless_count"] >= 10
+                if obj["motionless_count"] >= detect_config.stationary.threshold
                 # and it isn't due for a periodic check
                 and (
                     detect_config.stationary.interval == 0
@@ -893,12 +903,42 @@ def process_frames(
                 consolidated_detections = get_consolidated_object_detections(
                     detected_object_groups
                 )
+                tracked_detections = [
+                    d
+                    for d in consolidated_detections
+                    if d[0] not in all_attribute_labels
+                ]
                 # now that we have refined our detections, we need to track objects
-                object_tracker.match_and_update(frame_time, consolidated_detections)
+                object_tracker.match_and_update(frame_time, tracked_detections)
             # else, just update the frame times for the stationary objects
             else:
                 object_tracker.update_frame_times(frame_time)
 
+        # group the attribute detections based on what label they apply to
+        attribute_detections = {}
+        for label, attribute_labels in attribute_label_map.items():
+            attribute_detections[label] = [
+                d for d in consolidated_detections if d[0] in attribute_labels
+            ]
+
+        # build detections and add attributes
+        detections = {}
+        for obj in object_tracker.tracked_objects.values():
+            attributes = []
+            # if the objects label has associated attribute detections
+            if obj["label"] in attribute_detections.keys():
+                # add them to attributes if they intersect
+                for attribute_detection in attribute_detections[obj["label"]]:
+                    if box_inside(obj["box"], (attribute_detection[2])):
+                        attributes.append(
+                            {
+                                "label": attribute_detection[0],
+                                "score": attribute_detection[1],
+                                "box": attribute_detection[2],
+                            }
+                        )
+            detections[obj["id"]] = {**obj, "attributes": attributes}
+
         # debug object tracking
         if False:
             bgr_frame = cv2.cvtColor(
@@ -981,7 +1021,7 @@ def process_frames(
                 (
                     camera_name,
                     frame_time,
-                    object_tracker.tracked_objects,
+                    detections,
                     motion_boxes,
                     regions,
                 )