Video Style Transfer With Reinforcement Learning

Acknowledgement: The codebase is build on DiffuseST to extract latent representation for each frame and perform encoding and decoding stage of the diffusion model. All code for the policy gradient training loop, the policy network architecture, reward calculation, loss functions, and analysis is ours.

The main implementation is located in the DiffuseST_rl directory.

Training

To train the model, run:

python train.py [your arguments here]

Evaluation

To evaluate the stylized output images, run:

python eval.py [your arguments here]

Baseline

To try the baseline method, run:

python run_baseline.py [your arguments here]