Hello, I'm new to this job, your code use frame image after cutting through the ready-made model to forecast, I want to ask if there code that can directly predict the input video ? Thanks!

Hello, Thank you very much for roviding the C3D version of pytorch. I am trying to extract features for another dataset, which contains around 1748 frames in one video. While doing so I am facing an issue as following:-
RuntimeError: CUDA error: an illegal memory access was encountered

My input size is after passing through the function get_sport_clip() is torch.Size([1, 3, 1748, 112, 112]) I was wondering if you could kindly assist me in solving this issue.
Thank you very much once again.

Hi!

Thank you so much for your great work! I have a issue when loading the pre-trained, even if just using your predict.py:

Traceback (most recent call last):
  File "predict.py", line 101, in <module>
    main()
  File "predict.py", line 81, in main
    net.load_state_dict(torch.load('c3d.pickle'))
  File "/home/ye/anaconda2/lib/python2.7/site-packages/torch/serialization.py", line 231, in load
    return _load(f, map_location, pickle_module)
  File "/home/ye/anaconda2/lib/python2.7/site-packages/torch/serialization.py", line 386, in _load
    deserialized_objects[key]._set_from_file(f, offset)
RuntimeError: unexpected EOF. The file might be corrupted.

It seems that the c3d.pickle may be corrupted. Could you check with that please?

Thank you so much! :)

Hi, Thanks for such a great work! I just want to ask that did you get the c3d + svm = 82% stated in the paper?

I could not. Maybe I am doing something wrong!

Hello! I'm try to use your model to train UCF101, but the accuracy is only 50 % .Have you train the model on UCF101 datasets ？Is the test set accurate to 80% or more?

Since I am trying to use the pretrained weights for a different dataset, I am trying to understand how to change the very last layer. However when I try to read the pickle file with only pickle.load() it only returns a integer. When I try to read the pickle file sequentially, it says that A load persistent id instruction was encountered, but no persistent_load function was specified.

Since I don't know how the pickle file was generated, can you tell me how to read the pickle file in python, so I know what I can modify for my purposes ?

Hi, thanks for your work. I'd like to know if there is another link to download the pretrained weights (Sports 1M), the link provided is unavailable in this moments. Thank you

Hi, thank you for the implementation of the C3D net. I'm currently trying to train the model on the ucf dataset. As they described in the paper, I choose several (10/12/14/16) frames from a clip, unsqueeze each on dimension 1, then concatenate all of them. So the input tensor has a torch.Size([3, 10/12/14/16, 224, 224]).

Then I got a error: '' File "/home/cxing95/anaconda3/lib/python3.6/site-packages/torch/nn/functional.py", line 371, in max_pool3d ret = torch._C._nn.max_pool3d(input, kernel_size, stride, padding, dilation, ceil_mode)

RuntimeError: Given input size: (512x1x14x14). Calculated output size: (512x0x8x8). Output size is too small at /opt/conda/conda-bld/pytorch_1524584710464/work/aten/src/THCUNN/generic/VolumetricDilatedMaxPooling.cu:105 '' I'm not sure if it raised because of the input size. But could you tell what's the input should be like to train or use this net. That will help a lot.

Thank you.

are the pre-trained weights already cuda tensors? just calling torch.load on c3d.pickle raises a THCudaCheck out of memory exception for me.

If they are, might I suggest bringing them back to the cpu, so people can load them if they already have the gpu busy.

Hi,

First, thank you very much for contributing this c3d implementation in pytorch! I had a question on the origin of the pretrained weights, did you obtain them by converting them from another source or by training the network yourself ?