Video processing issues about mtts-can HOT 6 CLOSED

For the issue 2, can you please print out the tensor shape before fedding into the model (before model.predict())? I think you did not preprocess correctly. Printing shapes will help you a lot.

from mtts-can.

For the issue 1, you will need to look into the shape of preprocessing. Please go to inference_preprocess.py to print out shapes and try to debug what is going on. The other thing is that you did not specify the sampling rate flag, i don't know if that is the reason causing your problem.

from mtts-can.

I want to process live video streams instead of static .mp4 files and was wondering if there's a way to auto-detect the sampling rate for the videos or keeping a default value for the sampling rate flag that would allow processing live videos for the most part?

from mtts-can.

I beleive you can use CV2 or something else to detect the sampling rate e.g., https://learnopencv.com/how-to-find-frame-rate-or-frames-per-second-fps-in-opencv-python-cpp/.

I think most of the smartphone's front cameras probably are 30fps though.

from mtts-can.

@gabdalf1819, as a user i will recommend using some face detection algorithm as it is obvious from the screenshot attached its detecting stuff other than face only. for your second query, you need to count fps manually, there is no function in open cv. check this stackoverflow query

from mtts-can.

Please move to https://github.com/ubicomplab/rPPG-Toolbox

We added support there.

Thanks!

from mtts-can.