[Detector Support]: rarely if ever detects my cat

[h3llb3nt]

Describe the problem you are having

setup system in kitchen to patrol counter for my cat who is mostly white with black spots. my frigate config hardly ever detects him but occasionally for a brief moment see him as a cat or sometimes even as a dog. it has NO issue detecting me and my family as people tho. at first i thought it was the coral TPU not working but im having the same issue running on my intel GPU.

Version

0.17.1-416a9b7

Frigate config file

mqtt:
  enabled: true
  host: 192.168.225.13
  user: mqtt-user
  password: *****

#working edgetpu setup with yolov9 config 2026-06-03
#detectors:
#  coral:
#    type: edgetpu
#    device: pci
  #ov_0:
  #  type: openvino
  #  device: GPU

#model:
#  model_type: yolo-generic
#  width: 320
#  height: 320
#  path: /config/model_cache/yolov9/yolov9-s-relu6-best_320_int8_edgetpu.tflite
#  labelmap_path: /config/model_cache/yolov9/labels-coco17.txt

detectors:
  ov:
    type: openvino
    device: GPU

model:
  width: 300
  height: 300
  input_tensor: nhwc
  input_pixel_format: bgr
  path: /openvino-model/ssdlite_mobilenet_v2.xml
  labelmap_path: /openvino-model/coco_91cl_bkgr.txt


ffmpeg:
  hwaccel_args: preset-intel-qsv-h264

detect:
  enabled: true

objects:
  track:
    - cat
    - person
    - dog
  filters:
    cat:
      min_score: 0.1 # i've tweaked this several times to try to get ANY detection
      threshold: 0.1 
    person:
      threshold: 0.8
    dog:
      min_score: 0.1 # similiar test value to get any detection of the cat
      threshold: 0.1

cameras:
  kitchen_tapo_cam:
    enabled: true

    ffmpeg:
      inputs:
        #high res stream
        - path: rtsp://TapoKitchen:*****@192.168.225.239:554/stream1
          roles:
            - record
        #low res stream
        - path: rtsp://TapoKitchen:*****@192.168.225.239:554/stream2
          roles:
            - detect

    mqtt:
      enabled: true
      timestamp: false
      bounding_box: true
      crop: true

    zones:
      kitchen_counter_zone:
        coordinates: 
          0.14,0.44,0.245,0.62,0.543,0.374,0.849,0.505,0.867,0.305,0.555,0.211,0.461,0.232,0.244,0.362
        inertia: 1
        loitering_time: 1
        friendly_name: Kitchen Counter ZONE
        objects:
          - cat
          - dog
      kitchen_floor_zone:
        coordinates: 
          0.544,0.379,0.858,0.511,0.763,1,0.42,1,0.26,0.786,0.085,0.76,0.045,0.625,0.176,0.509,0.244,0.625
        inertia: 1
        loitering_time: 1
        friendly_name: Kitchen Floor ZONE

    snapshots:
      enabled: true
      required_zones:
        - kitchen_counter_zone
        - kitchen_floor_zone
      retain:
        default: 3
        objects:
          cat: 3
          person: 3
          dog: 3

    motion:
      mask:
        - 0.356,0.045,0.356,0,0,0,0,0.046
        - 0.289,0.186,0.298,0.311,0.397,0.253,0.391,0.144
        - 0.027,0.237,0.102,0.186,0.15,0.541,0.083,0.602
      threshold: 32
      contour_area: 10
      improve_contrast: false

record:
  enabled: true
  alerts:
    retain:
      days: 3
  detections:
    retain:
      days: 3
  #motion:
  #    threshold: 36
  #    contour_area: 10
  #    improve_contrast: false

review:
  alerts:
    required_zones:
      - kitchen_counter_zone
      - kitchen_floor_zone

go2rtc:
  streams:
    tapokitchen:
      - rtsp://TapoKitchen:*****@192.168.225.239:554/stream2

notifications:
  enabled: true
  email: *****@*****.com

semantic_search:
  enabled: false
  model_size: small

face_recognition:
  enabled: false
  model_size: small

lpr:
  enabled: false

classification:
  bird:
    enabled: false

version: 0.17-0

docker-compose file or Docker CLI command

services:
  frigate:
    container_name: frigate
    privileged: true
    restart: unless-stopped
    stop_grace_period: 30s
    image: ghcr.io/blakeblackshear/frigate:stable
    shm_size: "256MB"
    cap_add:
      - CAP_PERFMON
    devices:
      - /dev/dri/renderD128:/dev/dri/renderD128 # for intel & amd hwaccel, needs to be updated for your hardware
      - /dev/dri/card0:/dev/dri/card0
      - /dev/dri:/dev/dri
      - /dev/apex_0:/dev/apex_0
    volumes:
      - /home/docker/frigate-container/config:/config
      - /home/docker/frigate-container/storage:/media/frigate
      - /home/docker/frigate-container/plugins/edgetpu_tfl.py:/opt/frigate/frigate/detectors/plugins/edgetpu_tfl.py:ro
      - type: tmpfs # 1GB In-memory filesystem for recording segment storage
        target: /tmp/cache
        tmpfs:
          size: 1000000000
    ports:
      - "8971:8971"
      - "8554:8554" # RTSP feeds
      - "5000:5000"

Relevant Frigate log output

2026-06-05 18:29:12.655806711  [INFO] Preparing Frigate...
2026-06-05 18:29:12.956970556  [INFO] Starting Frigate...
2026-06-05 18:29:15.393435400  [2026-06-05 18:29:15] frigate.util.config            INFO    : Checking if frigate config needs migration...
2026-06-05 18:29:15.422764398  [2026-06-05 18:29:15] frigate.util.config            INFO    : frigate config does not need migration...
2026-06-05 18:29:16.943792598  [2026-06-05 18:29:16] frigate.app                    INFO    : Starting Frigate (0.17.1-416a9b7)
2026-06-05 18:29:16.961343838  [2026-06-05 18:29:16] peewee_migrate.logs            INFO    : Starting migrations
2026-06-05 18:29:16.961686362  [2026-06-05 18:29:16] peewee_migrate.logs            INFO    : There is nothing to migrate
2026-06-05 18:29:16.979803448  [2026-06-05 18:29:16] frigate.app                    INFO    : Recording process started: 528
2026-06-05 18:29:16.987776916  [2026-06-05 18:29:16] frigate.app                    INFO    : Review process started: 532
2026-06-05 18:29:16.989646425  [2026-06-05 18:29:16] frigate.app                    INFO    : go2rtc process pid: 119
2026-06-05 18:29:17.134157769  [2026-06-05 18:29:17] frigate.app                    INFO    : Embedding process started: 551
2026-06-05 18:29:17.163276470  [2026-06-05 18:29:17] frigate.app                    INFO    : Output process started: 598
2026-06-05 18:29:17.200915812  [2026-06-05 18:29:17] frigate.camera.maintainer      INFO    : Camera processor started for kitchen_tapo_cam: 655
2026-06-05 18:29:17.248770830  [2026-06-05 18:29:17] frigate.camera.maintainer      INFO    : Capture process started for kitchen_tapo_cam: 714
2026-06-05 18:29:17.440165024  [2026-06-05 18:29:17] frigate.api.fastapi_app        INFO    : Starting FastAPI app
2026-06-05 18:29:17.554412135  [2026-06-05 18:29:17] frigate.api.fastapi_app        INFO    : FastAPI started

Install method

Docker Compose

Object Detector

OpenVino

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

No response

[inkeep-ai-assistant[bot]]

The detection issue you're experiencing with your cat is likely related to the model you're using rather than your configuration. The default OpenVINO model (ssdlite_mobilenet_v2) is a general-purpose COCO-based model that may not perform well for detecting cats, especially ones with unusual coloring like a mostly white cat with black spots.

Your configuration settings for min_score and threshold at 0.1 are already extremely permissive, so the model itself is likely not recognizing your cat with sufficient confidence to register as a detection at all.

Recommendations:

1. Consider Frigate+ models - These are trained on real security camera footage from the Frigate community and typically provide much better accuracy for animals compared to generic COCO models. In your config, you can use a Frigate+ model by setting the path like:

model:
  path: plus://<model_id>

(1)

2. Check the Debug view - Navigate to the camera in the UI, select "Debug" at the top, and enable "Motion boxes" to verify that motion is being detected when your cat moves. (2) This will help confirm whether the issue is with motion detection or object classification.

3. Verify detect stream resolution - Your low-res stream (stream2) may have insufficient resolution for reliable cat detection. Smaller animals require adequate pixel coverage to be recognized. The detect resolution should provide enough detail for the model to identify the object. (1)

4. Review motion mask placement - Ensure your motion masks aren't covering areas where your cat typically moves. Motion masks prevent motion detection in those regions, which could prevent object detection from running. (3)

The fact that person detection works well but cat detection doesn't strongly suggests the model's training data doesn't generalize well to your specific cat's appearance.

📚 Sources:

1. Full Reference Config
2. Getting Started
3. Masks

---

Have another question? Just tag @inkeep.

[NickM-27]

There are a couple different issues here:

1. at the beginning the we can see the region (green box) is tightly around the cat. The fact that it was not detected is a model failure.
2. For the rest of the video the regions are very large, so the cat is a small portion of the overall image. Realistically it should still work but the model is clearly not working well.

For starters, using YOLOv9 would definitely be a good first step.

[h3llb3nt]

i'll switch back to yolo9 which i was using before with the coral...

[NickM-27]

what CPU do you have? I'd suggest trying the YOLOv9 with the iGPU if it can handle it

[h3llb3nt]

its an intel Core i5-1235U with lris Xe Graphics

[h3llb3nt]

it was working! and now its not... i gotta take a break from this...

[h3llb3nt]

i know that it CAN detect the cat because ive seen it happen for a very very brief period.

ive tried my m.2 coral TPU, ive tried different yolo models, logging for the screen shot i posted in the debug view showed nothing out of the ordinary.

motion IS being detected when the cat enters the frame