No person detection (no boxes), camera motion works but cannot be used as PIR for light automation

[Superlaust]

Describe the problem you are having

I’m trying to use Frigate person detection as a replacement for camera PIR to trigger a light automation, but no detections ever occur.

Streams are working, recording works, MQTT works, and Frigate UI is reachable, but:

No green detection boxes appear

No person/car events are generated

As a result, my light automation never triggers

I have detect: enabled: true on the cameras and objects: track: [person, car] configured, but Frigate does not detect anything even when a person clearly walks in front of the camera.

I’m unsure whether this is a configuration issue, a go2rtc/RTSP issue, or a Frigate 0.16 behavior change.

Version

0.16.3

What browser(s) are you using?

No response

Frigate config file

record:
  enabled: true
  retain:
    days: 7
    mode: all

mqtt:
  enabled: true
  host: <MQTT_IP>
  port: 1883
  user: <MQTT_USER>
  password: <MQTT_PASSWORD>
  client_id: <MQTT_CLIENT_ID>
  topic_prefix: frigate

detectors:
  cpu1:
    type: cpu

objects:
  track:
    - person
    - car

go2rtc:
  streams:
    udhaeng_main:
      - rtsp://<CAM_USER>:<CAM_PASS>@<CAM_IP_1>:554/h264Preview_01_main
    udhaeng_sub:
      - rtsp://<CAM_USER>:<CAM_PASS>@<CAM_IP_1>:554/h264Preview_01_sub
    braendeskur_main:
      - rtsp://<CAM_USER>:<CAM_PASS>@<CAM_IP_2>:554/h264Preview_01_main
    braendeskur_sub:
      - rtsp://<CAM_USER>:<CAM_PASS>@<CAM_IP_2>:554/h264Preview_01_sub

cameras:
  udhaeng:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/udhaeng_main
          input_args: preset-rtsp-restream
          roles: [record]
        - path: rtsp://127.0.0.1:8554/udhaeng_sub
          input_args: preset-rtsp-restream
          roles: [detect]
    detect:
      enabled: true
      width: 640
      height: 360
      fps: 5
    snapshots:
      enabled: true
      bounding_box: true
      retain:
        default: 7

  braendeskur:
    ffmpeg:
      inputs:
        - path: rtsp://127.0.0.1:8554/braendeskur_main
          input_args: preset-rtsp-restream
          roles: [record]
        - path: rtsp://127.0.0.1:8554/braendeskur_sub
          input_args: preset-rtsp-restream
          roles: [detect]
    detect:
      enabled: true
      width: 640
      height: 360
      fps: 5
    snapshots:
      enabled: true
      bounding_box: true
      retain:
        default: 7

version: 0.16-0

Relevant Frigate log output

2026-01-10 19:34:04.527158587  [2026-01-10 20:34:04] ffmpeg.braendeskur.detect      ERROR   : [vf#0:0 @ 0x55fa6de6b5c0] Terminating thread with return code -5 (Input/output error)
2026-01-10 19:39:04.090957058  [2026-01-10 20:39:04] frigate.video                  ERROR   : braendeskur: Unable to read frames from ffmpeg process.
2026-01-10 19:39:04.091348102  [2026-01-10 20:39:04] frigate.video                  ERROR   : braendeskur: ffmpeg process is not running. exiting capture thread...
2026-01-10 19:39:04.548120604  [2026-01-10 20:39:04] watchdog.braendeskur           ERROR   : Ffmpeg process crashed unexpectedly for braendeskur.
2026-01-10 19:39:04.549085027  [2026-01-10 20:39:04] watchdog.braendeskur           ERROR   : The following ffmpeg logs include the last 100 lines prior to exit.
2026-01-10 19:39:04.549704717  [2026-01-10 20:39:04] ffmpeg.braendeskur.detect      ERROR   : [AVHWFramesContext @ 0x7face0079680] Failed to sync surface 0x12: 1 (operation failed).
2026-01-10 19:39:04.550238832  [2026-01-10 20:39:04] ffmpeg.braendeskur.detect      ERROR   : [hwdownload @ 0x7facf4003980] Failed to download frame: -5.
2026-01-10 19:39:04.551074992  [2026-01-10 20:39:04] ffmpeg.braendeskur.detect      ERROR   : [vf#0:0 @ 0x559621e04f00] Error while filtering: Input/output error
2026-01-10 19:39:04.551815978  [2026-01-10 20:39:04] ffmpeg.braendeskur.detect      ERROR   : [vf#0:0 @ 0x559621e04f00] Task finished with error code: -5 (Input/output error)
2026-01-10 19:39:04.552343857  [2026-01-10 20:39:04] ffmpeg.braendeskur.detect      ERROR   : [vf#0:0 @ 0x559621e04f00] Terminating thread with return code -5 (Input/output error)
2026-01-10 19:41:36.675742369  [2026-01-10 20:41:36] frigate.video                  ERROR   : braendeskur: Unable to read frames from ffmpeg process.
2026-01-10 19:41:36.675876530  [2026-01-10 20:41:36] frigate.video                  ERROR   : braendeskur: ffmpeg process is not running. exiting capture thread...
2026-01-10 19:41:44.567136664  [2026-01-10 20:41:44] watchdog.braendeskur           ERROR   : Ffmpeg process crashed unexpectedly for braendeskur.
2026-01-10 19:41:44.567706716  [2026-01-10 20:41:44] watchdog.braendeskur           ERROR   : The following ffmpeg logs include the last 100 lines prior to exit.
2026-01-10 19:41:44.568088750  [2026-01-10 20:41:44] ffmpeg.braendeskur.detect      ERROR   : [AVHWFramesContext @ 0x7f3c24079800] Failed to sync surface 0x14: 1 (operation failed).
2026-01-10 19:41:44.568782779  [2026-01-10 20:41:44] ffmpeg.braendeskur.detect      ERROR   : [hwdownload @ 0x7f3c34003980] Failed to download frame: -5.
2026-01-10 19:41:44.569248521  [2026-01-10 20:41:44] ffmpeg.braendeskur.detect      ERROR   : [vf#0:0 @ 0x563439fb6000] Error while filtering: Input/output error
2026-01-10 19:41:44.569666850  [2026-01-10 20:41:44] ffmpeg.braendeskur.detect      ERROR   : [vf#0:0 @ 0x563439fb6000] Task finished with error code: -5 (Input/output error)
2026-01-10 19:41:44.569991952  [2026-01-10 20:41:44] ffmpeg.braendeskur.detect      ERROR   : [vf#0:0 @ 0x563439fb6000] Terminating thread with return code -5 (Input/output error)

Relevant go2rtc log output

No go2rtc warnings or errors appear in the Frigate add-on logs.
Streams load in the UI, but detection still produces no boxes/events.

FFprobe output from your camera

FFprobe output not available. RTSP streams are working; issue is detection only.

Frigate stats

Frigate UI is reachable.
MQTT connected.
Cameras show live video.
Detect and record enabled.
No detection events generated (no boxes).

Operating system

Home Assistant OS

Install method

Home Assistant Add-on

Object Detector

CPU (no coral)

Network connection

Wired

Camera make and model

Reolink PoE cameras Models: RLC-520A (2 units) Streams: h264Preview_01_main / h264Preview_01_sub via RTSP

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

Issue started after switching to Frigate Full Access and making config changes.
Setup previously worked with the same cameras and RTSP paths.
Rebooting Home Assistant and restarting the Frigate add-on did not resolve the issue.

[inkeep-ai-assistant[bot]]

Your ffmpeg process is crashing repeatedly with hardware acceleration errors, which is preventing object detection from running. The logs show Failed to sync surface and hwdownload errors, indicating issues with hardware-accelerated decoding(1).

The core issue: Your ffmpeg process can't decode frames, so the detector never receives frames to analyze - hence no detection boxes appear.

Solutions to try:

1. Disable hardware acceleration by setting hwaccel_args to an empty preset or removing it entirely. Add this to your config:

ffmpeg:
  hwaccel_args: []

(1)

2. For Reolink cameras specifically, use the http-flv stream format instead of RTSP for better stability. Update your go2rtc streams:

go2rtc:
  streams:
    udhaeng_main:
      - "ffmpeg:http://<CAM_IP_1>/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=<CAM_USER>&password=<CAM_PASSWORD>#video=copy#audio=copy#audio=opus"
    udhaeng_sub:
      - "ffmpeg:http://<CAM_IP_1>/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=<CAM_USER>&password=<CAM_PASSWORD>"

(2)

3. Ensure HTTP is enabled in your Reolink camera's advanced network settings(2).

4. Verify your detect stream resolution matches your camera's actual sub-stream output. The 640x360 you specified should work for most Reolink sub-streams, but confirm this matches your camera's settings(2).

Once ffmpeg can successfully decode frames without crashing, object detection will begin working and you'll see detection boxes appear.

📚 Sources:

1. Full Reference Config
2. Camera Specific Configurations

---

Have another question? Just tag @inkeep.