Batch Detection With video input frames


I want to classify objects in the video frames with my classifier. I can already find the positions and the related bbox of the objects in frames but when I want to classify them with learner.predict() It takes a lot of time for a single frame (nearly 5-10 objects per frame). So I wanted to make a batch prediction with the cropped object images. I searched about it but I can only find that we can do batch prediction in fastai with the test data. Is there a way to make batch prediction with multiple object images during in video frames ? (without image save / read operations. Images are already in the buffer)


1 Like