Repository navigation
[Support]: 0.18.2 – frigate/<camera>/motion MQTT topic and motion debug boxes never fire on 7 of 9 cameras (object detection works) #24805
Replies: 4 comments
|
Your config is the most likely culprit here, specifically the YAML structure under Look at this block: The A second thing worth checking: masks can be toggled off at runtime via MQTT, and that state is independent of the config. Check Note on the camera name used in these topics: use the camera's On the MQTT behavior you're seeing: after motion is initially detected, For the per-camera inspection you asked about, you can read and write the live values directly:
Subscribing to these per camera will tell you whether the effective runtime values on the 7 affected cameras actually match what you believe is in the config, which is the fastest way to confirm whether the YAML is being parsed as you expect. For visual confirmation of where the signal is lost, the built-in Motion Tuner lets you adjust values live and immediately see the effect on what is detected as motion. Navigate to Settings → Camera configuration → Motion tuner and select an affected camera. You can also use Debug Replay to re-run the detection pipeline against a recorded clip from an affected camera. Start it from History → {camera} → Actions → Debug Replay, and use From Timeline or Custom to add padding so motion detection has time to settle into the scene. Regarding 📚 Sources: Have another question? Just tag @inkeep. |
|
This is not a Frigate bug. The AI is partially correct here. Your YAML indentation is fine, and motion can't be turned off while detection is enabled, so neither of those is the cause. Your motion sensors on those cameras aren't broken, they're most likely stuck ON. Frigate only publishes to the motion topic when the state changes, and OFF is only sent after 30 seconds with no motion at all. If something in the frame changes constantly, like an on-screen timestamp/clock overlay, motion never goes OFF, so you never see another ON. That matches your HA sensor sitting "on" since startup. Do those 7 cameras have a timestamp overlay? If so, add a motion mask over it (like you have on porta_rua) or disable the overlay on the camera. The Motion Tuner in Settings will show you the boxes it's picking up. |
|
Thanks, that was it. What I did: added a motion mask over the clock on each camera (Settings > Camera configuration > Masks / Zones > Motion Mask). It applied live without a restart, and all 9 binary_sensor.*_motion now cycle on/off normally in Home Assistant when I walk in front of each camera. Two things that may help others:
I confirmed it with the Motion tuner, where a red box appeared intermittently on the seconds digits with nobody in view. I had earlier assumed the debug view showed no motion boxes at all, so that was my mistake. One thing I can't fully explain: one camera's motion sensor stayed "off" for about 35 hours while person events were still being created. After adding the masks everything has been consistent, so I'm treating this as not a Frigate bug. Feel free to close. Huge thank you for the help! |
|
Thank you!
you nailed it!
this was a great help to solve the problem.
im going to close the issue/hep request with the following
"Thanks, that was it. Every one of my 9 cameras has a burned-in date/time
overlay, and it was being picked up as constant motion, so the motion topic
never reached OFF and HA showed the sensors stuck "on".
What I did: added a motion mask over the clock on each camera (Settings >
Camera configuration > Masks / Zones > Motion Mask). It applied live
without a restart, and all 9 binary_sensor.*_motion now cycle on/off
normally in Home Assistant when I walk in front of each camera.
Two things that may help others:
- The mask has to cover the whole clock, including the edge pixels. On two
cameras it only worked after I enlarged the mask by a pixel or two.
- A zone is not enough: zones don't remove an area from motion detection,
only a motion mask does.
I confirmed it with the Motion tuner, where a red box appeared
intermittently on the seconds digits with nobody in view. I had earlier
assumed the debug view showed no motion boxes at all, so that was my
mistake.
One thing I can't fully explain: one camera's motion sensor stayed "off"
for about 35 hours while person events were still being created. After
adding the masks everything has been consistent, so I'm treating this as
not a Frigate bug. Feel free to close."
Huge thank you!
Luis Burnay
…On Mon, 26 Oct 2026 at 17:41, Josh Hawkins ***@***.***> wrote:
This is not a Frigate bug.
The AI is partially correct here. Your YAML indentation is fine, and
motion can't be turned off while detection is enabled, so neither of those
is the cause. Your motion sensors on those cameras aren't broken, they're
most likely stuck ON. Frigate only publishes to the motion topic when the
state changes, and OFF is only sent after 30 seconds with no motion at all.
If something in the frame changes constantly, like an on-screen
timestamp/clock overlay, motion never goes OFF, so you never see another
ON. That matches your HA sensor sitting "on" since startup. Do those 7
cameras have a timestamp overlay? If so, add a motion mask over it (like
you have on porta_rua) or disable the overlay on the camera. The Motion
Tuner in Settings will show you the boxes it's picking up.
—
Reply to this email directly, view it on GitHub
<#24805?email_source=notifications&email_token=BGNSFSJIYCBR6BCPBRSG3H35V6EFHA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOJSGAYTENRZUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVRTG633UMVZF6Y3MNFRWW#discussioncomment-19201269>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/BGNSFSM4TUYSU3ZAQUTKNRT5V6EFHAVCNFSNUABIKJSXA33TNF2G64TZHMYTMNZWHE2DCOJUHNCGS43DOVZXG2LPNY5TCMJRGM2DIMZYUF3AE>
.
Triage notifications, keep track of coding agent tasks and review pull
requests on the go with GitHub Mobile for iOS
<https://github.com/notifications/mobile/ios/BGNSFSMWAQOW7JC7ZEQBLF35V6EFHA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOJSGAYTENRZUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVJTG633UMVZF62LPOM>
and Android
<https://github.com/notifications/mobile/android/BGNSFSMQBQDBUEEI7UW5CBL5V6EFHA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOJSGAYTENRZUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVZTG633UMVZF6YLOMRZG62LE>.
Download it today!
You are receiving this because you authored the thread.Message ID:
***@***.***
com>
--
Luis Burnay
|
Uh oh!
There was an error while loading. Please reload this page.
Checklist
Describe the problem you are having
On 9 cameras with identical
motionsettings, only 2 (rampa_portao, porta_rua) publish frigate//motion (ON/OFF) and show motion boxes in the Debug view. The other 7 (jardim_tras, cozinha, escritorio, garagem, sala_lareira, deck_frente, rampa_lavandaria) never publish on frigate//motion, and the Debug view with "Motion boxes" enabled never shows a box, even while a person walks in front of the camera.Object detection on the same cameras works: person events are created (with clips), appear in Review, and "Bounding boxes" in Debug shows the person. _all_occupancy toggles correctly. Only the motion signal is missing.
Home Assistant (Frigate integration v5.15.6) shows binary_sensor.*_motion stuck: it goes off -> on for a few seconds right after a Frigate restart (that ON was not visible in an MQTT listener on frigate/+/motion), then never changes again. Example: jardim_tras motion on since the 0.18.2 start (13:37 UTC) and unchanged when a person walked past at 14:54 UTC, while all_occupancy went off at 14:54:10.
Timeline: last real ON for jardim_tras was 2026-10-24 23:52:51 UTC, 6 s after a container start. Since then only the restart-time ON. Happened on 0.18.1 and still on 0.18.2 (upgrade did not change it). I don't know whether it worked on 0.16/0.17.
also,
the docs AI listed skip_motion_threshold, lightning_threshold, masks and thresholds; all identical or unset on all cameras.”
Steps to reproduce
Version
0.18.2-cfa08c5
In which browser(s) are you experiencing the issue with?
chrome on windows and on android phone
Frigate config file
docker-compose file or Docker CLI command
Relevant Frigate log output
Relevant go2rtc log output
No output in the last 30 minutes (no warnings or errors). go2rtc runs as a separate container (alexxit/go2rtc:latest) serving all streams on :8554.Operating system
Proxmox
Install method
Proxmox via Docker
Network connection
Wired
Camera make and model
escritorio: Hikvision DS-2CD2420F-IW (works: no, motion silent) cozinha, sala_lareira: same fisheye model (motion silent) porta_rua: different brand (motion works) (fill in the models of the other cameras if you know them)
Screenshots of the Frigate UI's System metrics pages
Screenshots not attached: cameras are all at ~5 fps with 0 skipped frames, CPU detector ~10 ms inference; see "Any other information".
Any other information that may be helpful
Stream comparison (real sub stream via ffprobe; detect is scaled to 640x480 on every camera):
No correlation with resolution or fps.
Full per-camera config diff between a working (rampa_portao) and broken (jardim_tras) camera: only paths, stream names and zones differ; the generated ffmpeg commands are identical.
Zones: the 2 working cameras have a zone, but deck_frente has one and is still broken. Adding a zone to jardim_tras (jardim_tras_zona1) made no difference.
Checked with no effect: restarting Frigate; upgrading 0.18.1 -> 0.18.2; reloading the HA integration; motion settings identical on all 9 (enabled, threshold 30, contour_area 10, lightning_threshold 0.8, skip_motion_threshold null, improve_contrast true, mqtt_off_delay 30). Only porta_rua has a motion mask.
Question: is there a way to dump per-camera motion state or enable motion debug logging to see where the signal is lost?
All reactions