This PR adds a second recording mode to SSL-Vision. This mode records frames (.png) and accompanying detection frames (.json) nonstop, and immediately writes these to disk. An optional (and enabled by default) FPS limiter will prevent the PC from being flooded with write operations.
The collection of this data could help develop SSL-Vision. For example, the data could be used to train models for automatic color calibration, as well as adaptive color calibration when lighting conditions change.
While the previous version introduced a slight stutter in the robot detection thread, this version runs all writing in a separate thread (thanks to @mamiksik) and does not have any performance impact (as far as I have measured).
This PR adds a second recording mode to SSL-Vision. This mode records frames (.png) and accompanying detection frames (.json) nonstop, and immediately writes these to disk. An optional (and enabled by default) FPS limiter will prevent the PC from being flooded with write operations.
The collection of this data could help develop SSL-Vision. For example, the data could be used to train models for automatic color calibration, as well as adaptive color calibration when lighting conditions change.
While the previous version introduced a slight stutter in the robot detection thread, this version runs all writing in a separate thread (thanks to @mamiksik) and does not have any performance impact (as far as I have measured).