Wei Jiang (Columbia University), “Advanced Techniques for Semantic Concept Detection in General Videos” (2009)

Wei Jiang (Columbia University), “Advanced Techniques for Semantic Concept Detection in General Videos” (2010). Advisor: Prof. Shih-Fu Chang

The thesis investigates semantic concept detection in general videos focusing on two advanced directions: multi-concept learning and multi-modality learning. For multi-concept learning, we provide an in-depth study of jointly detecting multiple concepts in general domains, where concept relations are hard to compute. Specifically, we propose two methods with the frameworks of two-layer Context-Based Concept Fusion (CBCF) and single-layer multi-label classification, respectively.  These methods use semantic context information to enhance detection of individual concepts.  For multi-modality learning, we conduct an in-depth investigation of combining audio and visual cues to help detect generic concepts. We develop methods with two strategies: global fusion of features or models from audio and visual modalities, and construction of the local audio-visual atomic representation to enforce a moderate-level audio-visual synchronization. Audio and visual modalities complement each other and their simultaneous processing can uncover relations that are otherwise unavailable when considering the modalities separately.

Click here to access the thesis or contact the author.