search

MichiganCOG / ViP

Video Platform for Action Recognition and Object Detection in Pytorch
MIT License
219 stars 37 forks source link

RuntimeError: Given input size: (512x1x7x7). Calculated output size: (512x0x4x4). Output size is too small #47

Closed MrWwei closed 4 years ago

MrWwei commented 4 years ago

I try to use shorter clips to train. eg.15

natlouis commented 4 years ago

If you are using C3D, that architecture requires minimum clip length of 16 frames.

You can see our pre-processing options here: https://github.com/MichiganCOG/ViP/blob/master/datasets/abstract_datasets.py#L61 By default, we loop clips that are below the minimum requirement.