Disentangling Controllable Object Through Video Prediction Improves Visual Reinforcement Learning.
Yuanyi ZhongAlexander G. SchwingJian PengPublished in: ICASSP (2020)
Keyphrases
- reinforcement learning
- visual objects
- visual input
- visual cues
- prediction accuracy
- visual data
- visual information
- real time
- visual appearance
- visual analysis
- visual properties
- video streams
- object detection and tracking
- video data
- video sequences
- video database
- video content
- spatial relations
- d objects
- low level
- object tracking
- video search
- function approximation
- learning algorithm
- object motion
- category specific
- news video
- object model
- visual features
- multi agent
- multimedia
- bounding box
- video analysis
- reinforcement learning algorithms
- video surveillance
- video frames
- space time
- vision system
- object detection
- moving objects
- object recognition
- stationary camera
- combining information from multiple