Learning Semantic-Aware Dynamics for Video Prediction
This work addresses video prediction for applications like autonomous systems or video editing, but it appears incremental as it builds on existing decomposition and inpainting techniques.
The paper tackles video frame prediction by modeling dis-occlusions and semantically consistent regions, decomposing scene layout and motion into layers for prediction and fusion, resulting in a model that explicitly represents objects and learns class-specific motion.
We propose an architecture and training scheme to predict video frames by explicitly modeling dis-occlusions and capturing the evolution of semantically consistent regions in the video. The scene layout (semantic map) and motion (optical flow) are decomposed into layers, which are predicted and fused with their context to generate future layouts and motions. The appearance of the scene is warped from past frames using the predicted motion in co-visible regions; dis-occluded regions are synthesized with content-aware inpainting utilizing the predicted scene layout. The result is a predictive model that explicitly represents objects and learns their class-specific motion, which we evaluate on video prediction benchmarks.