| Technical Name |
Deep Reinforcement Learning for Video Prediction |
| Project Operator |
National Chiao Tung University |
| Project Host |
杭學鳴 |
| Summary |
Our video prediction approach uses sparse motion field which combines deep RL technique to discover the critical sparse positionstheir motion vectors. We apply POBMC method for final frame synthesis. Our prediction shows similar PSNRSSIM score compared with SOTA but much better human vision perceptual LPIPS score.
|
| Scientific Breakthrough |
Our model combines video compression POBMC techniquemodern DLRL. Unlike direct generation, our model explicitly models the motion tendencies. Different from dense method which requires artificial constraint, ours preserves the true motions. Besides, our model size is smallwe only need fewer data to achieve good generalization.
|
| Industrial Applicability |
From industry application perspective, video prediction can be used in any application that needs to predict the future, such as smart autonomous vehicles. Or video prediction can benefit other tasks by predicting the future first such as activities predictions. This method can also be applied on video compression. We only need to transmit a small amount of motion vectorstheir location inform |
| Keyword |
Video prediction Deep learning Reinforcement learning Sparse motion vector Overlapped blocked motion compensation |
