Multimodal Compact Bilinear Pooling for Visual Question Answering and Visual Grounding

Akira Fukui, Dong Huk Park, Daylen Yang, Anna Rohrbach, Trevor Darrell, Marcus Rohrbach

(Submitted on 6 Jun 2016 (v1), last revised 24 Sep 2016 (this version, v3))

Modeling textual or visual information with vector representations trained from large language or visual datasets has been successfully explored in recent years. However, tasks such as visual question answering require combining these vector representations with each other. Approaches to multimodal pooling include element-wise product or sum, as well as concatenation of the visual and textual representations. We hypothesize that these methods are not as expressive as an outer product of the visual and textual vectors. As the outer product is typically infeasible due to its high dimensionality, we instead propose utilizing Multimodal Compact Bilinear pooling (MCB) to efficiently and expressively combine multimodal features. We extensively evaluate MCB on the visual question answering and grounding tasks. We consistently show the benefit of MCB over ablations without MCB. For visual question answering, we present an architecture which uses MCB twice, once for predicting attention over spatial features and again to combine the attended representation with the question representation. This model outperforms the state-of-the-art on the Visual7W dataset and the VQA challenge.

Comments:Accepted to EMNLP 2016Subjects:Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)Cite as:arXiv:1606.01847 [cs.CV] (or arXiv:1606.01847v3 [cs.CV] for this version)

Submission history

From: Marcus Rohrbach [view email] 
[v1] Mon, 6 Jun 2016 17:59:56 GMT (2194kb,D)
[v2] Thu, 23 Jun 2016 19:52:41 GMT (3358kb,D)
[v3] Sat, 24 Sep 2016 01:58:59 GMT (3443kb,D)