When calculating FVD, FID and IS scores, how many fake videos (sample.mp4 ) need to be generated? Whether you use all real video frames when calculating these scores?
CUDA_VISIBLE_DEVICES=gpu_id python ./src/scripts/calc_metrics_for_dataset.py
--fake_data_path /path/to/video/frames
--real_data_path /path/to/dataset/frames
Look forward to your reply.