Handle back seeking going to previous clip

* perf(track): avoid numpy reductions on tiny box lists in position smoothing

update_position runs per tracked object per frame. While a position has
fewer than 10 samples it calls np.percentile four times, and average_boxes
(per stationary object per frame) calls np.mean four times - all on lists of
at most 10 ints, where numpy's per-call dispatch/validation overhead
dominates the actual work.

Replace them with pure-Python equivalents:
- average_boxes: sum()/len() instead of np.mean (bit-identical output)
- interpolated_percentile(): linear-interpolated percentile matching
  numpy.percentile (including its lerp branch at frac>=0.5) for the small
  lists used here, in place of np.percentile

Measured in the release image (numpy 1.26.4) on a 10-element list:
np.percentile 18735 ns -> 191 ns/call (98x); np.mean-based average_boxes
7480 ns -> 591 ns (12.7x); ~74 us saved per object-frame in update_position.
A live py-spy --gil profile of a camera process_frames worker showed
np.percentile (update_position) and np.mean (average_boxes) among the top
Frigate-owned on-CPU frames.

Output is unchanged: added tests assert both helpers are bit-identical to
numpy over randomized small inputs.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* Drop interpolated_percentile, keep only average_boxes

Per review: reimplementing np.percentile hurts readability and risks
divergence from numpy (e.g. numpy 2.x). Revert update_position to
np.percentile and remove the helper; keep only the average_boxes change
(sum()/len() instead of np.mean), which stays bit-identical.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

docs/docs/configuration/face_recognition.md

@@ -86,7 +86,7 @@ Navigate to <NavPath path="Settings > Enrichments > Face recognition" />.
 - **Detection threshold**: Face detection confidence score required before recognition runs. This field only applies to the standalone face detection model; `min_score` should be used to filter for models that have face detection built in.
   - Default: `0.7`
 - **Minimum face area**: Minimum size (in pixels) a face must be before recognition runs. Depending on the resolution of your camera's `detect` stream, you can increase this value to ignore small or distant faces.
-  - Default: `500` pixels
+  - Default: `750` pixels
 
 </TabItem>
 <TabItem value="yaml">
@@ -95,7 +95,7 @@ Navigate to <NavPath path="Settings > Enrichments > Face recognition" />.
 face_recognition:
   enabled: true
   detection_threshold: 0.7
-  min_area: 500
+  min_area: 750
 ```
 
 </TabItem>

docs/docs/configuration/metrics.md

@@ -59,6 +59,8 @@ Metrics are available at `/api/metrics` by default. No additional Frigate config
 - `frigate_storage_used_bytes{storage=""}` - Storage used bytes
 - `frigate_storage_mount_type{mount_type="", storage=""}` - Storage mount type info
 
+These gauges report the operating system's figures for the whole filesystem (the same numbers as `df`), not Frigate's own recording footprint. For how this differs from the recordings usage shown in the UI, see [Understanding storage usage](/configuration/record#understanding-storage-usage).
+
 ### Service Metrics
 
 - `frigate_service_uptime_seconds` - Uptime in seconds

docs/docs/configuration/record.md

@@ -199,10 +199,6 @@ Because recording segments are written in 10 second chunks, pre-capture timing d
 
 Pre and post capture footage is included in the **recording timeline**, visible in the History view. Note that pre/post capture settings only affect which recording segments are **retained on disk** — they do not change the start and end points shown in the UI. The History view will still center on the review item's actual time range, but you can scrub backward and forward through the retained pre/post capture footage on the timeline. The Explore view shows object-specific clips that are trimmed to when the tracked object was actually visible, so pre/post capture time will not be reflected there.
 
-## Will Frigate delete old recordings if my storage runs out?
-
-If there is less than an hour left of storage, the oldest hour of recordings will be deleted and a message will be printed in the Frigate logs. This emergency cleanup deletes the oldest recordings first regardless of retention settings to reclaim space as quickly as possible.
-
 ## Configuring Recording Retention
 
 Frigate supports both continuous and tracked object based recordings with separate retention modes and retention periods.
@@ -355,3 +351,63 @@ Setting `verbose: true` writes a detailed report of every orphaned file and data
 This operation uses considerable CPU resources and includes a safety threshold that aborts if more than 50% of files would be deleted. Only run when necessary. If you set `force: true` the safety threshold will be bypassed; do not use `force` unless you are certain the deletions are intended.
 
 :::
+
+## Understanding storage usage
+
+The storage usage Frigate reports will not exactly match what the operating system reports with `df` or `du`. This is expected, not a bug. The sections below explain how Frigate derives its storage figures and why they differ from the disk's own accounting.
+
+### How Frigate measures recording usage
+
+The **Recordings** value on the Storage Metrics page (<NavPath path="System > Storage" />) — and the per-camera **Camera Storage** breakdown — is the sum of the recording segment sizes Frigate has written, taken from Frigate's database. It is **not** computed by a scan of the disk. Frigate tracks usage this way by design: repeatedly walking the entire drive to total its size would keep hard drives spun up and add unnecessary I/O.
+
+The disk **total** shown beside it, and the free-space figure Frigate uses to decide when to delete recordings, instead come from the operating system's report for the whole filesystem mounted at `/media/frigate`. As a result, the **Unused** value on the page is _total disk capacity minus Frigate's recordings_ — not the drive's real free space, which will be lower whenever anything else is stored on the disk.
+
+### What counts toward usage — and why it won't match `df`
+
+Only **recording segments** (`/media/frigate/recordings`) are included in the recordings storage total. Plenty of other things consume real disk space but are **not** part of that number:
+
+- **Snapshots and thumbnails** (`/media/frigate/clips`) — see [Snapshots](/configuration/snapshots). These are retained independently of recordings.
+- **Preview videos** and **review thumbnails** (also under `/media/frigate/clips`).
+- **Exports** (`/media/frigate/exports`) — exports are never removed by retention.
+- **The database, downloaded detection models, and face / license plate training images** (stored under `/config`).
+- **Debug images from enrichments** (`/media/frigate/clips`) — when enabled, License Plate Recognition's `debug_save_plates` and GenAI's `debug_save_thumbnails` save plate crops and request images for troubleshooting.
+
+These files are the usual explanation for an "other" or seemingly unaccounted bucket of space — it is real, it is Frigate's, and it simply isn't part of the _recordings_ total. They are also why comparing the **Recordings** figure to `df -h` always shows a gap: `df` additionally counts any non-Frigate data on the disk, filesystem overhead and reserved blocks (ext4 reserves ~5% for root by default, so a disk can read "full" before recordings approach the total), and recently deleted recordings whose space has not yet been reclaimed.
+
+:::tip
+
+The Storage page is not intended to be a system-wide disk monitor — it shows how much space _Frigate's recordings_ use. To see true disk usage, use `df -h` (free space) and `du -sh` (per-directory usage) on the host.
+
+:::
+
+### Free space and the `/media/frigate` mount
+
+Frigate reports the capacity and free space of whatever filesystem is actually mounted at `/media/frigate` **inside the container**. If an external drive or network share isn't truly mounted there — a missing `/etc/fstab` entry, a share that was offline when the container started, or a host that doesn't pass the path through — the container falls back to the host's OS disk, and Frigate will correctly report that smaller disk instead of the drive you intended.
+
+If the reported capacity doesn't match your drive, the mount is the place to look, not Frigate. Verify what is actually mounted from inside the container:
+
+```bash
+docker exec -it frigate df -h /media/frigate
+docker exec -it frigate mount | grep media
+```
+
+See the [storage mount layout](/frigate/installation#storage) for how the volumes are expected to be configured.
+
+### The `/tmp/cache` area is separate
+
+Recording segments are first written to `/tmp/cache` — a small, in-memory (`tmpfs`) area — before being checked and moved to `/media/frigate/recordings`. Because it is separate and small, `/tmp/cache` can fill up and produce `No space left on device` errors even when the recordings disk has plenty of room — they are different storage areas. See [Recordings troubleshooting](/troubleshooting/recordings) for diagnosing cache and slow-storage issues.
+
+### When the metrics don't match what's on disk
+
+Because usage is tracked in the database, deleting recording files directly on disk — or files left behind after an upgrade — will not update the reported usage, and can even push it above 100%. Frigate is unaware of files it didn't record and won't count or remove them automatically. Use [Syncing Media Files With Disk](#syncing-media-files-with-disk) to reconcile the database with what is actually on disk.
+
+## Will Frigate delete old recordings if my storage runs out?
+
+Yes. Frigate continuously checks the **free space of the disk** holding `/media/frigate/recordings`. This is different from adding up the size of every recording: free space is a single number the operating system already tracks, so Frigate can ask for it instantly without reading through your files or spinning up the disk — which is exactly why it relies on this check rather than scanning the drive. When less than roughly one hour of recording space remains — estimated from the current recording bitrate, **not** a fixed percentage — Frigate deletes the oldest recordings to reclaim space and logs a message. This emergency cleanup removes the oldest recordings first **regardless of retention settings**.
+
+Two consequences follow from this being based on whole-disk free space:
+
+- Because the check uses the disk's real free space, **anything** filling the drive — including non-Frigate files — can trigger deletion of your oldest recordings.
+- Cleanup can run while a meaningful percentage of the disk is still free (for example, with high bitrates or many cameras), because the threshold is "less than ~1 hour of recording headroom," not "X% full."
+
+Frequent emergency cleanups usually mean your configured retention exceeds what the disk can hold. Reduce your retention days so the normal retention cleanup keeps up and the emergency path rarely triggers.

docs/docs/frigate/planning_setup.md

@@ -42,6 +42,8 @@ Frigate requires a CPU with AVX + AVX2 instructions. Most modern CPUs (post-2011
 
 Storage is an important consideration when planning a new installation. To get a more precise estimate of your storage requirements, you can use an IP camera storage calculator. Websites like [IPConfigure Storage Calculator](https://calculator.ipconfigure.com/) can help you determine the necessary disk space based on your camera settings.
 
+Once running, see [Understanding storage usage](/configuration/record#understanding-storage-usage) for how Frigate measures and reports disk usage — and why its numbers won't exactly match `df` or `du`.
+
 #### SSDs (Solid State Drives)
 
 SSDs are an excellent choice for Frigate, offering high speed and responsiveness. The older concern that SSDs would quickly "wear out" from constant video recording is largely no longer valid for modern consumer and enterprise-grade SSDs.

frigate/record/maintainer.py

@@ -42,6 +42,8 @@ from frigate.util.services import get_video_properties
 
 logger = logging.getLogger(__name__)
 
+STALE_RECORDINGS_INFO_TTL = MAX_SEGMENTS_IN_CACHE * MAX_SEGMENT_DURATION * 2
+
 
 class SegmentInfo:
     def __init__(
@@ -301,6 +303,8 @@ class RecordingMaintainer(threading.Thread):
                 RecordingsDataTypeEnum.saved.value,
             )
 
+        self._expire_stale_recordings_info(grouped_recordings)
+
         recordings_to_insert: list[Optional[dict[str, Any]]] = await asyncio.gather(
             *tasks
         )
@@ -311,6 +315,21 @@ class RecordingMaintainer(threading.Thread):
             [r for r in recordings_to_insert if r is not None],
         )
 
+    def _expire_stale_recordings_info(
+        self, grouped_recordings: defaultdict[str, list[dict[str, Any]]]
+    ) -> None:
+        expire_before = datetime.datetime.now().timestamp() - STALE_RECORDINGS_INFO_TTL
+        for recordings_info in (
+            self.object_recordings_info,
+            self.audio_recordings_info,
+        ):
+            for camera in list(recordings_info.keys()):
+                if camera in grouped_recordings:
+                    continue
+                info = recordings_info[camera]
+                while info and info[0][0] < expire_before:
+                    info.pop(0)
+
     def drop_segment(self, cache_path: str) -> None:
         Path(cache_path).unlink(missing_ok=True)
         self.end_time_cache.pop(cache_path, None)

frigate/test/test_builtin.py

@@ -0,0 +1,41 @@
+"""Tests for frigate.util.builtin helpers."""
+
+import unittest
+from unittest.mock import patch
+
+from frigate.util.builtin import EventsPerSecond
+
+
+class TestEventsPerSecond(unittest.TestCase):
+    def test_eps_is_zero_before_any_events(self) -> None:
+        eps = EventsPerSecond()
+        with patch("frigate.util.builtin.time.monotonic", return_value=100.0):
+            self.assertEqual(eps.eps(), 0.0)
+
+    def test_eps_counts_events_in_window(self) -> None:
+        eps = EventsPerSecond(last_n_seconds=10)
+        clock = [1000.0]
+        with patch("frigate.util.builtin.time.monotonic", side_effect=lambda: clock[0]):
+            eps.start()
+            # one event per second for five seconds
+            for _ in range(5):
+                clock[0] += 1.0
+                eps.update()
+            # five events over the five seconds since start
+            self.assertAlmostEqual(eps.eps(), 1.0)
+
+    def test_old_timestamps_expire_from_window(self) -> None:
+        eps = EventsPerSecond(last_n_seconds=10)
+        clock = [0.0]
+        with patch("frigate.util.builtin.time.monotonic", side_effect=lambda: clock[0]):
+            eps.start()
+            for _ in range(10):
+                clock[0] += 1.0
+                eps.update()
+            # jump well past the window so every timestamp ages out
+            clock[0] += 100.0
+            self.assertEqual(eps.eps(), 0.0)
+
+
+if __name__ == "__main__":
+    unittest.main()

frigate/test/test_maintainer.py

@@ -115,6 +115,46 @@ class TestMaintainer(unittest.IsolatedAsyncioTestCase):
         self.assertIsNone(result)
         maintainer.drop_segment.assert_called_once_with(cache_path)
 
+    async def test_expire_stale_recordings_info_drops_only_absent_cameras(self):
+        config = MagicMock(spec=FrigateConfig)
+        config.cameras = {}
+        stop_event = MagicMock()
+        maintainer = RecordingMaintainer(config, stop_event)
+
+        now = datetime.datetime.now().timestamp()
+        ancient = now - 86400
+        recent = now - 1
+
+        maintainer.object_recordings_info["present_cam"] = [(ancient, [], [], [])]
+        maintainer.audio_recordings_info["present_cam"] = [(ancient, 0, [])]
+
+        maintainer.object_recordings_info["absent_cam"] = [
+            (ancient, [], [], []),
+            (recent, [], [], []),
+        ]
+        maintainer.audio_recordings_info["absent_cam"] = [
+            (ancient, 0, []),
+            (recent, 0, []),
+        ]
+
+        grouped_recordings = {"present_cam": [{"start_time": ancient}]}
+
+        maintainer._expire_stale_recordings_info(grouped_recordings)
+
+        self.assertEqual(
+            maintainer.object_recordings_info["present_cam"], [(ancient, [], [], [])]
+        )
+        self.assertEqual(
+            maintainer.audio_recordings_info["present_cam"], [(ancient, 0, [])]
+        )
+
+        self.assertEqual(
+            maintainer.object_recordings_info["absent_cam"], [(recent, [], [], [])]
+        )
+        self.assertEqual(
+            maintainer.audio_recordings_info["absent_cam"], [(recent, 0, [])]
+        )
+
 
 if __name__ == "__main__":
     unittest.main()

frigate/test/test_obects.py

@@ -1,6 +1,22 @@
+import random
 import unittest
 
+import numpy as np
+
 from frigate.track.tracked_object import TrackedObjectAttribute
+from frigate.util.object import average_boxes
+
+
+class TestBoxStatistics(unittest.TestCase):
+    def test_average_boxes_matches_numpy(self) -> None:
+        rng = random.Random(0)
+        for _ in range(5000):
+            boxes = [
+                [rng.randint(0, 4000) for _ in range(4)]
+                for _ in range(rng.randint(1, 10))
+            ]
+            expected = [float(np.mean([b[i] for b in boxes])) for i in range(4)]
+            self.assertEqual(average_boxes(boxes), expected)
 
 
 class TestAttribute(unittest.TestCase):

frigate/test/test_stationary_classifier.py

@@ -0,0 +1,39 @@
+"""Tests for stationary object classification thresholds."""
+
+import unittest
+
+from frigate.track.stationary_classifier import (
+    DEFAULT_OBJECT_THRESHOLDS,
+    DYNAMIC_OBJECT_THRESHOLDS,
+    NON_STATIONARY_OBJECT_THRESHOLDS,
+    STATIONARY_OBJECT_THRESHOLDS,
+    StationaryThresholds,
+    get_stationary_threshold,
+)
+
+
+class TestStationaryThresholds(unittest.TestCase):
+    def test_known_labels_return_expected_singletons(self) -> None:
+        self.assertIs(get_stationary_threshold("package"), STATIONARY_OBJECT_THRESHOLDS)
+        self.assertIs(get_stationary_threshold("car"), DYNAMIC_OBJECT_THRESHOLDS)
+        self.assertIs(
+            get_stationary_threshold("license_plate"),
+            NON_STATIONARY_OBJECT_THRESHOLDS,
+        )
+
+    def test_unknown_label_returns_shared_default(self) -> None:
+        # an unknown label must reuse the shared default instance, not allocate
+        # a fresh one on every call (this runs per object per frame)
+        first = get_stationary_threshold("person")
+        second = get_stationary_threshold("dog")
+        self.assertIs(first, DEFAULT_OBJECT_THRESHOLDS)
+        self.assertIs(second, DEFAULT_OBJECT_THRESHOLDS)
+
+    def test_default_matches_a_fresh_instance(self) -> None:
+        # the shared default must be value-equivalent to the previous
+        # per-call StationaryThresholds()
+        self.assertEqual(DEFAULT_OBJECT_THRESHOLDS, StationaryThresholds())
+
+
+if __name__ == "__main__":
+    unittest.main()

frigate/test/test_video.py

@@ -82,6 +82,24 @@ class TestRegion(unittest.TestCase):
 
         assert len(cluster_candidates) == 2
 
+    def test_cluster_candidates_partition_boxes(self):
+        # every box index must appear in exactly one cluster (no box used twice,
+        # none dropped) - the invariant the used-box tracking enforces
+        boxes = [
+            (100, 100, 200, 200),
+            (202, 150, 252, 200),
+            (210, 160, 260, 210),
+            (900, 900, 950, 950),
+            (905, 905, 955, 955),
+        ]
+
+        cluster_candidates = get_cluster_candidates(
+            self.frame_shape, self.min_region_size, boxes
+        )
+
+        assigned = [idx for cluster in cluster_candidates for idx in cluster]
+        self.assertEqual(sorted(assigned), list(range(len(boxes))))
+
     def test_transliterate_to_latin(self):
         self.assertEqual(transliterate_to_latin("frégate"), "fregate")
         self.assertEqual(transliterate_to_latin("utilité"), "utilite")

frigate/track/norfair_tracker.py

@@ -641,9 +641,11 @@ class NorfairTracker(ObjectTracker):
             self.deregister(self.track_id_map[e_id], e_id)
 
         # update list of object boxes that don't have a tracked object yet
-        tracked_object_boxes = [obj["box"] for obj in self.tracked_objects.values()]
+        tracked_object_boxes = {
+            tuple(obj["box"]) for obj in self.tracked_objects.values()
+        }
         self.untracked_object_boxes = [
-            o[2] for o in detections if o[2] not in tracked_object_boxes
+            o[2] for o in detections if tuple(o[2]) not in tracked_object_boxes
         ]
 
     def print_objects_as_table(self, tracked_objects: Sequence) -> None:

frigate/track/stationary_classifier.py

@@ -63,6 +63,9 @@ NON_STATIONARY_OBJECT_THRESHOLDS = StationaryThresholds(
     max_stationary_history=4,
 )
 
+# Default thresholds for any other object label
+DEFAULT_OBJECT_THRESHOLDS = StationaryThresholds()
+
 
 def get_stationary_threshold(label: str) -> StationaryThresholds:
     """Get the stationary thresholds for a given object label."""
@@ -76,7 +79,7 @@ def get_stationary_threshold(label: str) -> StationaryThresholds:
     if label in NON_STATIONARY_OBJECT_THRESHOLDS.objects:
         return NON_STATIONARY_OBJECT_THRESHOLDS
 
-    return StationaryThresholds()
+    return DEFAULT_OBJECT_THRESHOLDS
 
 
 class StationaryMotionClassifier:

frigate/util/builtin.py

@@ -2,7 +2,6 @@
 
 import ast
 import copy
-import datetime
 import logging
 import math
 import multiprocessing.queues
@@ -10,7 +9,9 @@ import queue
 import re
 import shlex
 import struct
+import time
 import urllib.parse
+from collections import deque
 from collections.abc import Mapping
 from multiprocessing.managers import ValueProxy
 from pathlib import Path
@@ -32,23 +33,20 @@ class EventsPerSecond:
         self._start = None
         self._max_events = max_events
         self._last_n_seconds = last_n_seconds
-        self._timestamps = []
+        self._timestamps: deque[float] = deque(maxlen=max_events)
 
     def start(self) -> None:
-        self._start = datetime.datetime.now().timestamp()
+        self._start = time.monotonic()
 
     def update(self) -> None:
-        now = datetime.datetime.now().timestamp()
+        now = time.monotonic()
         if self._start is None:
             self._start = now
         self._timestamps.append(now)
-        # truncate the list when it goes 100 over the max_size
-        if len(self._timestamps) > self._max_events + 100:
-            self._timestamps = self._timestamps[(1 - self._max_events) :]
         self.expire_timestamps(now)
 
     def eps(self) -> float:
-        now = datetime.datetime.now().timestamp()
+        now = time.monotonic()
         if self._start is None:
             self._start = now
         # compute the (approximate) events in the last n seconds
@@ -63,7 +61,7 @@ class EventsPerSecond:
     def expire_timestamps(self, now: float) -> None:
         threshold = now - self._last_n_seconds
         while self._timestamps and self._timestamps[0] < threshold:
-            del self._timestamps[0]
+            self._timestamps.popleft()
 
 
 class InferenceSpeed:

frigate/util/object.py

@@ -339,18 +339,13 @@ def reduce_boxes(boxes, iou_threshold=0.0):
 
 def average_boxes(boxes: list[list[int, int, int, int]]) -> list[int, int, int, int]:
     """Return a box that is the average of a list of boxes."""
-    x_mins = []
-    y_mins = []
-    x_max = []
-    y_max = []
-
-    for box in boxes:
-        x_mins.append(box[0])
-        y_mins.append(box[1])
-        x_max.append(box[2])
-        y_max.append(box[3])
-
-    return [np.mean(x_mins), np.mean(y_mins), np.mean(x_max), np.mean(y_max)]
+    n = len(boxes)
+    return [
+        sum(box[0] for box in boxes) / n,
+        sum(box[1] for box in boxes) / n,
+        sum(box[2] for box in boxes) / n,
+        sum(box[3] for box in boxes) / n,
+    ]
 
 
 def median_of_boxes(boxes: list[list[int, int, int, int]]) -> list[int, int, int, int]:
@@ -401,13 +396,13 @@ def get_cluster_candidates(frame_shape, min_region, boxes):
     # determined by the max_region size minus half the box + 20%
     # TODO: see if we can do this with numpy
     cluster_candidates = []
-    used_boxes = []
+    used_boxes = set()
     # loop over each box
     for current_index, b in enumerate(boxes):
         if current_index in used_boxes:
             continue
         cluster = [current_index]
-        used_boxes.append(current_index)
+        used_boxes.add(current_index)
         cluster_boundary = get_cluster_boundary(b, min_region)
         # find all other boxes that fit inside the boundary
         for compare_index, compare_box in enumerate(boxes):
@@ -436,7 +431,7 @@ def get_cluster_candidates(frame_shape, min_region, boxes):
 
             if should_cluster:
                 cluster.append(compare_index)
-                used_boxes.append(compare_index)
+                used_boxes.add(compare_index)
         cluster_candidates.append(cluster)
 
     # return the unique clusters only
@@ -558,6 +553,7 @@ def reduce_detections(
                 current_detection = sorted_by_area[current_detection_idx]
                 current_label = current_detection[0]
                 current_box = current_detection[2]
+                current_area = area(current_box)
                 overlap = 0
                 for to_check_idx in range(
                     min(current_detection_idx + 1, len(sorted_by_area)),
@@ -568,14 +564,14 @@ def reduce_detections(
                     # if area of current detection / area of check < 5% they should not be compared
                     # this covers cases where a large car parked in a driveway doesn't block detections
                     # of cars in the street behind it
-                    if area(current_box) / area(to_check) < 0.05:
+                    if current_area / area(to_check) < 0.05:
                         continue
 
                     intersect_box = intersection(current_box, to_check)
                     # if % of smaller detection is inside of another detection, consolidate
-                    if intersect_box is not None and area(intersect_box) / area(
-                        current_box
-                    ) > LABEL_CONSOLIDATION_MAP.get(
+                    if intersect_box is not None and area(
+                        intersect_box
+                    ) / current_area > LABEL_CONSOLIDATION_MAP.get(
                         current_label, LABEL_CONSOLIDATION_DEFAULT
                     ):
                         overlap = 1

web/src/components/player/HlsVideoPlayer.tsx

@@ -47,6 +47,7 @@ type HlsVideoPlayerProps = {
   frigateControls?: boolean;
   inpointOffset?: number;
   onClipEnded?: (currentTime: number) => void;
+  onClipPrevious?: (diff: number) => void;
   onPlayerLoaded?: () => void;
   onTimeUpdate?: (time: number) => void;
   onPlaying?: () => void;
@@ -74,6 +75,7 @@ export default function HlsVideoPlayer({
   frigateControls = true,
   inpointOffset = 0,
   onClipEnded,
+  onClipPrevious,
   onPlayerLoaded,
   onTimeUpdate,
   onPlaying,
@@ -339,11 +341,17 @@ export default function HlsVideoPlayer({
           onSeek={(diff) => {
             const currentTime = videoRef.current?.currentTime;
 
-            if (!videoRef.current || !currentTime) {
+            if (!videoRef.current || currentTime == undefined) {
               return;
             }
 
-            videoRef.current.currentTime = Math.max(0, currentTime + diff);
+            const newTime = currentTime + diff;
+
+            if (newTime < 0 && onClipPrevious) {
+              onClipPrevious(diff);
+            } else {
+              videoRef.current.currentTime = Math.max(0, newTime);
+            }
           }}
           onSetPlaybackRate={(rate) => {
             setPlaybackRate(rate, true);

web/src/components/player/dynamic/DynamicVideoPlayer.tsx

@@ -49,6 +49,7 @@ type DynamicVideoPlayerProps = {
   onControllerReady: (controller: DynamicVideoController) => void;
   onTimestampUpdate?: (timestamp: number) => void;
   onClipEnded?: () => void;
+  onClipPrevious?: (diff: number) => void;
   onSeekToTime?: (timestamp: number, play?: boolean) => void;
   setFullResolution: React.Dispatch<React.SetStateAction<VideoResolutionType>>;
   toggleFullscreen: () => void;
@@ -68,6 +69,7 @@ export default function DynamicVideoPlayer({
   onControllerReady,
   onTimestampUpdate,
   onClipEnded,
+  onClipPrevious,
   onSeekToTime,
   setFullResolution,
   toggleFullscreen,
@@ -343,6 +345,7 @@ export default function DynamicVideoPlayer({
           onTimeUpdate={onTimeUpdate}
           onPlayerLoaded={onPlayerLoaded}
           onClipEnded={onValidateClipEnd}
+          onClipPrevious={onClipPrevious}
           onSeekToTime={(timestamp, play) => {
             if (onSeekToTime) {
               onSeekToTime(timestamp, play);

web/src/views/recording/RecordingView.tsx

@@ -336,6 +336,16 @@ export function RecordingView({
     [currentTimeRange, updateSelectedSegment],
   );
 
+  const onClipPrevious = useCallback(
+    (diff: number) => {
+      manuallySetCurrentTime(
+        currentTime + diff,
+        mainControllerRef.current?.isPlaying() ?? false,
+      );
+    },
+    [currentTime, manuallySetCurrentTime],
+  );
+
   const onShareReviewLink = useCallback(
     (timestamp: number) => {
       const reviewUrl = createRecordingReviewUrl(location.pathname, {
@@ -902,6 +912,7 @@ export function RecordingView({
                     );
                   }}
                   onClipEnded={onClipEnded}
+                  onClipPrevious={onClipPrevious}
                   onSeekToTime={manuallySetCurrentTime}
                   onControllerReady={(controller) => {
                     mainControllerRef.current = controller;