Modeling Interval Temporal Dependencies for Complex Activities Understanding
Abstract
Complex activity typically consists of temporally sequential or overlapping primitive events occurring over a time interval. The existing dynamic models are point-based and they cannot effectively model event temporal dependences. To overcome this limitation, we introduce the Interval Temporal Bayesian Network (ITBN), a novel graphical model that combines the Bayesian Network with the Interval Algebra, to explicitly model the temporal dependencies over time intervals. Furthermore, to handle the challenge with explicit primitive event detection and tracking in real world videos, we propose to use topic models to perform implicit event detection. Combining ITBN model with the topic models yields a powerful framework that can perform complex activity recognition without explicit primitive event detection and tracking. The proposed framework is evaluated on two computer vision applications: human body activity recognition and human facial activity recognition.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 11, 2013
- Accession Number
- ADA607523
Entities
People
- Qiang Ji
Organizations
- Rensselaer Polytechnic Institute