Can you verify the logic behind confidence on this one? Getting 1.0 confidence on anything I put on the screen (1.0 confidence by itself should be extremely rare). Thanks.

@PaulTR I understand. I will check it

Hi @PaulTR ,
I'm a new member and will be working with you in the near future.

I checked the result of the last frame, and found that the result always returns a confidence score of 1.0 as shown in the attached image.

So I took that exact model, plugged it into our MediaPipe sample, and then added the score into the logs to make sure it's something the model can generate. Something is missing in the LiteRT app for extracting the confidences, but it should be there (potentially that first line in your log, but the order is wrong for it being 12.399% confident and classifying as '7' when there's a 14.2% earlier).