On-device object training support within Frigate

[Shreyas-R]

Describe what you are trying to accomplish and why in non technical terms Need on-device object training support within frigate to reduce object detection errors and to train custom objects which are not available in the default model used by Frigate.

Describe the solution you'd like It would be great if Frigate supports,

1. Basic tensorflow lite model with on device training support.
2. UI within frigate to locally train and untrain object models (something similar to double-take used for face recognition).

Describe alternatives you've considered Frigate+ is coming which is cloud based solution. I'm looking for something offline and local solution. The following are the other alternatives but they have their own disadvantages.

To reduce object detection errors:

1. Apply object detection filters tweaking min_score, threshold and mask.
2. Train custom tensorflow lite models outside Frigate and use them within Frigate.

To train custom objects:

1. Train custom tensorflow lite models outside Frigate and use them within Frigate.

Additional context Raising this feature request after facing lot of false detections in Frigate which is discussed here #3792.

[blakeblackshear]

I don't think this will ever be implemented directly in Frigate. I could see it as a companion service alongside Frigate like double take, but my time and energy will be going into Frigate+.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.