论文信息 - EURECOM at TRECVID 2017: The Adhoc Video Search

EURECOM at TRECVID 2017: The Adhoc Video Search

This paper describes the submissions of the EURECOM team to the TRECVID 2017 AVS task. Our approach is to project both the text topic and the visual keyframes in the same vector space, corresponding to a word embedding. We compare and combine several word embeddings, also using specific keyword weightings.

Bernard Mérialdo | Danny Francis | Benoit Huet

[1] Andrew Zisserman,et al. Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[2] Marco Idiart,et al. Matrix Factorization using Window Sampling and Negative Sampling for Improved Word Representations , 2016, ACL.

[3] Pietro Perona,et al. Microsoft COCO: Common Objects in Context , 2014, ECCV.

[4] Wenpeng Yin,et al. Learning Word Meta-Embeddings , 2016, ACL.

[5] Fei-Fei Li,et al. Deep visual-semantic alignments for generating image descriptions , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[6] Xueqi Cheng,et al. Learning Word Representations by Jointly Modeling Syntagmatic and Paradigmatic Relations , 2015, ACL.

[7] Benoit Huet,et al. EURECOM at TRECVID 2016: The Adhoc Video Search and Video Hyperlinking Tasks , 2016, TRECVID.

[8] Georges Quénot,et al. TRECVID 2017: Evaluating Ad-hoc and Instance Video Search, Events Detection, Video Captioning and Hyperlinking , 2017, TRECVID.

[9] Dennis Koelma,et al. The ImageNet Shuffle: Reorganized Pre-training for Video Event Detection , 2016, ICMR.