AVA results visualization

Hi!

I have been running the test on the AVA dataset. It works perfectly and returns the expected mAP as stated in the model ZOO files. My question now is if there is any way to visualize the results on the input video/frames, either live while the testing is being performed or after it has finished, in a similar way to your demo gif. I understand that this is to be done with the bounding boxes from _ava_detection_val_boxes_and_labels.csv_ file and the predictions generated while running the test in the _detections_latest.csv_ file.

By exploring a bit the code, I have found the __images_and_boxes_preprocessing_cv2_ in the _ava_dataset.py_ file: https://github.com/facebookresearch/SlowFast/blob/master/slowfast/datasets/ava_dataset.py#L99-L208
which I understand is made for this purpose. This function is called in function getitem: https://github.com/facebookresearch/SlowFast/blob/master/slowfast/datasets/ava_dataset.py#L337-L413
Are we supposed to call this last function for visualization in _test_net.py_ somehow? Or these functions are not in any way relevant for this purpose? And as the _detections_latest.csv_ contains the prediction for all the listed actions for each detected person, which method would you recommend to only display the relevant ones?

Thank you for sharing your code!
Best regards!