TokyoTechCanon at TRECVID 2013

We aim at developing a high-performance system using Gaussian-mixture-model (GMM) supervectors and tree-structured GMMs [6, 7, 8] for the semantic indexing task [1, 2, 3, 4]. GMM supervectors corresponding to six types of audio and visual features are extracted from video shots. Tree-structured GMMs reduce the computational cost of maximum a posteriori (MAP) adaptation for estimating GMM parameters while keeping accuracy at high levels. This year, we improve our re-scoring method using video-clip scores by introducing a scaling parameter. Here, the video-clip score is defined as the maximum value of shot scores among all the shots in a video clip. Our best result was 28.4% in terms of Mean InfAP, which was ranked third among participating teams in the semantic indexing task.