About extract features

[IVparagigm]

Hi, I try to use extract_features.py to get image feature, the dataset I used is Charades. Should I use rgb_charades.pt as the pre-trained model in extract_features.py? Another question is why in train_i3d.py, the transfrom has videotransforms.RandomHorizontalFlip() but in extract_features.py transfrom is only transforms.Compose([videotransforms.CenterCrop(224)])
Thank you if you can help me.