Inconsistent detections on Raspberry Pi 5 with Hailo 8L

General
,
1
  • We are running our model on USB camera stream in a car. The point is that, we get inconsistent detections for people in the back seat, although, people in the front seat get consistent detections.
  • We have used Hailo’s ByteTrack but we don’t get consistent tracking ID either.

What are the options we have?

2

Welcome to the Hailo Community!

Do you get inconsistent detections with the model in floating point format? In that case you may want to check the original model training. The lighting conditions are likely different and you may need to add additional training data.

If there is a large degradation between the floating point model and the converted model you should look into the advanced optimization tutorials in the AI Software suite Docker.

hailo tutorials

This will open a Jupyter notebook server with tutorials for each step of the model conversion process including advanced optimizations.

1 Like