Du Tran

change this line https://github.com/facebook/C3D/blob/master/C3D-v1.1/src/caffe/layers/deconvolution3d_layer.cpp#L89 to top_shape[0] = bottom[0]->shape(0); may solve it.

Also it is good to share your full error log, something before and after F0430 07:16:29.406836 3839 blob.hpp:140] Check failed: num_axes() 4 axes.

@spiritotaku pull and try again.

you can reduce lr or stop training. Is this UCf101?

Why do you use sampling rate is too high? e.g. 25

1 mean that you keep all frames, 2 mean that you skip every other frames, and so on. Now, if your video has 30fps, then your reader will read at...

it will resize your images to height x width resolution.

@prateek-kacker have you resolved your issue?

Yes, we can do that if there are enough interest.

We are going to release a new version C3D together with, R3D, R(2+1)D [https://128.84.21.199/pdf/1711.11248.pdf] in caffe2 which is pretty much similar to pyTorch. Hope it will be helpful to you!