Knowledge Transfer for Scene-specific Motion Prediction
This work addresses motion prediction in video analysis for applications like surveillance or autonomous systems, but it is incremental as it builds on prior knowledge integration without introducing a fundamentally new approach.
The paper tackles the problem of predicting scene-specific motion patterns from a single video frame by exploiting the interplay between agent dynamics and scene semantics, introducing a Dynamic Bayesian Network that uses patch descriptors to encode movement probabilities, and demonstrates accurate trajectory prediction with the ability to transfer predictions to novel scenes with similar elements.
When given a single frame of the video, humans can not only interpret the content of the scene, but also they are able to forecast the near future. This ability is mostly driven by their rich prior knowledge about the visual world, both in terms of (i) the dynamics of moving agents, as well as (ii) the semantic of the scene. In this work we exploit the interplay between these two key elements to predict scene-specific motion patterns. First, we extract patch descriptors encoding the probability of moving to the adjacent patches, and the probability of being in that particular patch or changing behavior. Then, we introduce a Dynamic Bayesian Network which exploits this scene specific knowledge for trajectory prediction. Experimental results demonstrate that our method is able to accurately predict trajectories and transfer predictions to a novel scene characterized by similar elements.