Time-Contrastive Networks
An pytorch implementation a time-contrastive networks model as presented in the paper "Time-Contrastive Networks: Self-Supervised Learning from Multi-View Observation" (Sermanet Et al. 2017).
Only training from single view observations has been implemented. Videos are added to the data/train and data/validation directories.
Training the model is done by running the train_tcn.py script.