I’m testing transfer learning based on a ResNext3D model using a custom dataset.
I have two questions:
Is the same frame rate required in my dataset? THe UCF101 human action videos uses 25 FPS (and 320×240 resolution). I used 10 frames per second until now to have a smaller total dataset size.
Is the ResNext3D model the best to use as a base for transfer learning in action recognition? Is there a model with better performance released?
Thanks a lot for any answers, links or tips!
 Original UCF101 Paper: https://arxiv.org/pdf/1212.0402.pdf