Frigate 0.16 beta FAQ

[hawkeye217]

This FAQ is maintained by the Frigate developers to help answer some of the common questions with Frigate 0.16 beta. If you have a question that's not answered here, be sure to read the release notes, the official documentation, and search the recent discussions first. If you still need help, open a support discussion.

Updated June 28, 2025

Upgrade issues

1. I just upgraded and Frigate is not detecting objects anymore. Why?
   This was discussed in the breaking changes in the release notes. Frigate 0.16 disables object detection by default. For most users, a config migrator is run to enable detection but if this fails or your config is read-only, you may need to edit your config to enable detection.

   detect:
     enabled: true
   

Usage

1. I enabled Face Recognition but I see the message WARNING : camera is configured to track ['face'] objects, which are not supported by the current model. in my logs. Why?
   This warning appears when you’ve added face (or license_plate) to the objects --> track list in your config, but you're not using a model that supports those object types, such as Frigate+ or a custom model with native support. If you're running the default free model that ships with Frigate, Face Recognition and License Plate Recognition will still work without those objects in the track list. To remove the log warning, simply delete face and/or license_plate from objects --> track. See the model requirements in the Face Recognition and LPR documentation.

2. How do I use detected faces and license plates in Home Assistant sensors and automations?
   The Frigate HA Integration version 5.9.2 adds support for a sensor for each camera's last recognized face / license plate. For more advanced used cases, Frigate updates the frigate/events and frigate/tracked_object_update MQTT topics with recognized faces and license plates. You can create MQTT sensors or use these topics directly in automations. See the documentation. Here is an example automation for faces that can be adapted for license plates easily.

   alias: Face Recognition Notification
   description: "Send notification when a face is recognized by Frigate"
   trigger:
     - platform: mqtt
       topic: frigate/tracked_object_update 
   condition:
     - condition: template
       value_template: "{{ trigger.payload_json.type == 'face' }}"
     - condition: template
       value_template: "{{ trigger.payload_json.name != 'unknown' }}"
     - condition: template
       value_template: "{{ trigger.payload_json.score > 0.7 }}"
   action:
     - service: notify.mobile_app_your_device
       data:
         title: "Face Detected: {{ trigger.payload_json.name }}"
         message: >-
           {{ trigger.payload_json.name }} detected on {{ trigger.payload_json.camera | replace('_', ' ') | title }}
           Confidence: {{ (trigger.payload_json.score * 100) | round }}%
   

3. I've set the large model in my config but how can I be sure enrichments are using my iGPU? The usage in System Metrics is 0.
   There is a bug in Intel's GPU tools that prevents correct reporting of the usage to Frigate. You can run intel_gpu_top on your host to see GPU usage.

4. I am using a Coral or Hailo for object detection, can I use it for enrichments like face recognition and LPR?
   No, a Coral is a TPU (tensor processing unit). Frigate requires a dedicated GPU (graphics processing unit) like an Intel iGPU or Nvidia graphics card to accelerate enrichments with hardware. If you have both a Coral or Hailo and a dedicated GPU, it is possible to still use the Coral for object detection. In all cases, however, you will need to use the appropriate Docker image for your installation to accelerate enrichments. See the documentation.

5. Frigate loses connection to my cameras and then consumes a lot of memory until it becomes unresponsive. Why?
   Users with this issue are often running WiFi cameras, which are not recommended. A few users have reported that this issue can be mitigated by using preset-rtsp-generic instead of preset-rtsp-restream. But in general, this is usually a result of WiFi cameras or is an ffmpeg issue that only happens with certain cameras.

6. I'm having trouble using the mask/zone editor, I can't click points and close the polygon.
   This is due to Brave Browser's underlying extreme anti-fingerprinting code that breaks a common function in JavaScript. To use Brave Browser with Frigate, you should disable Brave Shields.

7. Can I use the record stream for LPR and Face Recognition? I don't get good results with the detect stream.
   No, your record stream is not decoded by Frigate. Like other enrichments, LPR and Face Recognition run on the stream defined with the detect role in your config. However, when using the recommended cameras, we found through extensive testing that a 1280x720 stream was sufficient for most users for accurate recognition. You could either:

   • Use the main stream from your camera as-is for the detect role, or
   • Use the main stream for the detect role but set your detect width and height to 1280 and 720 respectively so that your GPU will be used to scale it.

8. Why does my Intel GPU usage always show as 0? Is my iGPU being used?
   We've received many reports of this where users indicate that their Intel GPU stats work for a while, but then eventually show as zero. This is a bug in Intel GPU Tools, there's nothing Frigate can do to fix it at this point.