CMU-informedia @ TRECViD 2014 semantic indexing
Lu Jiang, Xiaojun Chang, Zexi Mao, Anil Armagan, Zhengzhong Lan, Xuanchong Li, Shoou-I Yu, Yi Ping Yang, Deyu Meng, Pinar Duygulu-Sahin, Alexander G. Hauptmann · Monash University Research Portal (Monash University) · 2014
We report on our system used in the TRECVID 2014 Semantic Indexing (SIN) task. We highlight the following new components: 1) self-paced learning pipeline for concept training, 2) dense trajectory with fisher vector encoding, 3) multi-modal pseudo relevance feedback for final results reranking and 4) deep convolutional neural networks directly trained on SIN keyframes. With the help of above components, we were ranked top 3 among all type A runs (using only TRECVID IACC training data).