Neural Multimodal Cooperative Learning Toward Micro-Video Understanding
- 1 July 2019
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 29 (10577149) , 1-14
Abstract
The prevailing characteristics of micro-videos result in the less descriptive power of each modality. The micro-video representations, several pioneer efforts proposed, are limited in implicitly exploring the consistency between different modality information but ignore the complementarity. In this paper, we focus on how to explicitly separate the consistent features and the complementary features from the mixed information and harness their combination to improve the expressiveness of each modality. Towards this end, we present a Neural Multimodal Cooperative Learning model (NMCL) to split the consistent component and the complementary component by a novel relation-aware attention mechanism. Specifically, the computed attention score can be used to measure the correlation between the features extracted from different modalities. And then, a threshold is learned for each modality to distinguish the consistent and complementary features, according to the score. Thereafter, we integrate the consistent parts to enhance the representations and supplement the complementary ones to reinforce the information in each modality. As to the problem of redundant information, which may cause overfitting and is hard to distinguish, we devise an attention network to dynamically capture the features which closely related the category and output a discriminative representation for prediction. Experimental results on a real-world micro-video dataset show that NMCL outperforms state-of-the-art methods. Further studies verify the effectiveness and cooperative effects brought by the attentive mechanism.Keywords
Funding Information
- National Basic Research Program of China (2015CB352502)
- National Natural Science Foundation of China (61772310, 61702300, 61702302)
- Project of Thousand Youth Talents 2016
- Tencent AI Lab Rhino-Bird Joint Research Program (JR201805)
- National Natural Science Foundation of China (61625301, 61731018)
- Microsoft Research
This publication has 37 references indexed in Scilit:
- Word Embedding based Generalized Language Model for Information Retrieval2015
- Cross-Platform Emerging Topic Detection and Elaboration from Multimedia StreamsACM Transactions on Multimedia Computing, Communications, and Applications, 2015
- A Review and Meta-Analysis of Multimodal Affect Detection SystemsACM Computing Surveys, 2015
- Retrieval-Based Face Annotation by Weak Label Regularized Local Coordinate CodingIEEE Transactions on Pattern Analysis and Machine Intelligence, 2013
- Multiview Metric Learning with Global Consistency and Local SmoothnessACM Transactions on Intelligent Systems and Technology, 2012
- Unsupervised Discovery of Mid-Level Discriminative Patches2012
- Modeling Latent Discriminative Dynamic of Multi-dimensional Affective Signals2011
- Large-Scale Parallel Collaborative Filtering for the Netflix Prize2008
- On Information and SufficiencyThe Annals of Mathematical Statistics, 1951
- RELATIONS BETWEEN TWO SETS OF VARIATESBiometrika, 1936