TSN-Pytorch

We have released MMAction, a full-fledged action understanding toolbox based on PyTorch. It includes implementation for TSN as well as other STOA frameworks for various tasks. The lessons we learned in this repo are incorporated into MMAction to make it bettter. We highly recommend you switch to it. This repo will remain here for historical references.

Note: always use git clone --recursive https://github.com/yjxiong/tsn-pytorch to clone this project.
Otherwise you will not be able to use the inception series CNN archs.

This is a reimplementation of temporal segment networks (TSN) in PyTorch. All settings are kept identical to the original caffe implementation.

For optical flow extraction and video list generation, you still need to use the original TSN codebase.

Training

To train a new model, use the main.py script.

The command to reproduce the original TSN experiments of RGB modality on UCF101 can be

bashCopy
python main.py ucf101 RGB <ucf101_rgb_train_list> <ucf101_rgb_val_list> \
   --arch BNInception --num_segments 3 \
   --gd 20 --lr 0.001 --lr_steps 30 60 --epochs 80 \
   -b 128 -j 8 --dropout 0.8 \
   --snapshot_pref ucf101_bninception_ 

For flow models:

bashCopy
python main.py ucf101 Flow <ucf101_flow_train_list> <ucf101_flow_val_list> \
   --arch BNInception --num_segments 3 \
   --gd 20 --lr 0.001 --lr_steps 190 300 --epochs 340 \
   -b 128 -j 8 --dropout 0.7 \
   --snapshot_pref ucf101_bninception_ --flow_pref flow_  

For RGB-diff models:

bashCopy
python main.py ucf101 RGBDiff <ucf101_rgb_train_list> <ucf101_rgb_val_list> \
   --arch BNInception --num_segments 7 \
   --gd 40 --lr 0.001 --lr_steps 80 160 --epochs 180 \
   -b 128 -j 8 --dropout 0.8 \
   --snapshot_pref ucf101_bninception_ 

Testing

After training, there will checkpoints saved by pytorch, for example ucf101_bninception_rgb_checkpoint.pth.

Use the following command to test its performance in the standard TSN testing protocol:

bashCopy
python test_models.py ucf101 RGB <ucf101_rgb_val_list> ucf101_bninception_rgb_checkpoint.pth \
   --arch BNInception --save_scores <score_file_name>

Or for flow models:

bashCopy
python test_models.py ucf101 Flow <ucf101_rgb_val_list> ucf101_bninception_flow_checkpoint.pth \
   --arch BNInception --save_scores <score_file_name> --flow_pref flow_