bellos1203/tcd

Code for "Class-Incremental Learning for Action Recognition in Videos", ICCV 2021

tcd's Issues

Txt file

'ucf101/splits/ucf101_rgb_train_split_1.txt

where do you get the text file.
after i extract the frame how can i generate the label for UCF101 dataset.

out of memory(CPU)

I have encountered a problem where exemplar selection step consumes a lot of memory(CPU) but that(GPU)，and the server with the memory of 64GB is completely insufficient.

How much memory is appropriate？

run.sh

Hello,

Upon reviewing the project to reproduce the experiments, I noticed that the run.sh file is missing.

Question about test_crops setting

Hello, thanks for your wonderful work. I have two question about test_crops
1、Why the test_crops is set as 5. The commen setting of test_crops are 1 in action recognition.
2、Do all experiments of all datasets are conducted in the condition of test_crops = 5?

Waiting for your kindly reply.

sthv2 dataset error

Traceback (most recent call last):
File "/home3/ls/TCD/main.py", line 83, in
main()
File "/home3/ls/TCD/main.py", line 58, in main
train_i_cl.train_task(args, i, current_task, current_head, class_indexer, prefix=prefix)
File "/home3/ls/TCD/train/train_i_cl.py", line 339, in train_task
tcd.update_importance(args,model,train_loader,criterion)
File "/home3/ls/TCD/cl_methods/tcd.py", line 21, in update_importance
loss.backward()
File "/home/user/anaconda3/envs/ls/lib/python3.10/site-packages/torch/_tensor.py", line 487, in backward
torch.autograd.backward(
File "/home/user/anaconda3/envs/ls/lib/python3.10/site-packages/torch/autograd/init.py", line 197, in backward
Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
File "/home/user/anaconda3/envs/ls/lib/python3.10/site-packages/torch/nn/modules/module.py", line 61, in call
raise RuntimeError("You are trying to call the hook of a dead Module!")
RuntimeError: You are trying to call the hook of a dead Module!

out