Representing Videos based on Scene Layouts for Recognizing Agent-in-Place Actions.
Ruichi YuHongcheng WangAng LiJingxiao ZhengVlad I. MorariuLarry S. DavisPublished in: CoRR (2018)
Keyphrases
- atomic actions
- video sequences
- human actions
- multiple agents
- action selection
- dynamic scenes
- decision theoretic
- action recognition
- action detection
- cognitive agents
- human object interactions
- video scene
- human activities
- motion features
- agent behavior
- action primitives
- plan library
- goal directed
- multi agent
- video images
- multi agent systems
- neighboring frames
- moving camera
- rational agents
- video sequences captured
- multiagent systems
- learning agent
- plan recognition
- d scene
- autonomous agents
- reward function
- input video
- traffic scenes
- intelligent agents
- interacting agents
- partial observations
- partially observable
- external events
- software agents
- video frames
- scene analysis
- current situation
- visual data
- agent architecture
- world model
- spatio temporal
- three dimensional
- joint action
- dynamic environments
- simulated robot
- moving objects
- space time interest points
- video content
- temporal structure
- state transitions
- activity recognition
- mobile agents
- plan execution
- dynamic textures
- past experience
- video surveillance
- agent receives
- image sequences
- situation calculus
- video data
- state action
- single image
- markov decision processes
- input image
- reinforcement learning
- sensing actions
- multiagent reinforcement learning