This repository contains the code for the CS7180 course project: Action Segmentation for Instructional Videos with Semi-Weak Supervision.
The data is available on raptor on /mnt/raptor/yuhan/Breakfast_per_task. It can be accessed via Cygnus server.
The environment can be installed via anaconda:
conda create --name <env> --file requirements.txt
To train the model, please run the following command:
python joint_train.py
The default setting is training model on 50% weakly labeled videos with temporal consistency regularization, which does not use unlabeled videos.
To incorporate unlabeled videos for training, please run:
python joint_train.py --dwsa_type mean
or
python joint_train.py --dwsa_type min
For inference, please run:
python inference.py
or
python inference.py --dwsa_type (mean/min)
For evaluation, please run:
python eval.py --recog_dir results_split1
You may change the recog_dir if you save the results in a different folder.