Using pre-trained model for video feature extraction

Hi!
I wanted to use a pre-trained-X3D model for extracting features from a video by removing the final classification layer from the model. I have already extracted frames from the video but could not find support for running pre-trained models on custom set of frames. TIA