Solving VIsual Madlibs with Multiple Cues

Tatiana Tommasi, Arun Mallya, Bryan Plummer, Svetlana Lazebnik, Alexander Berg and Tamara Berg

Abstract

This paper focuses on answering fill-in-the-blank style multiple choice questions from the Visual Madlibs dataset. Previous approaches to Visual Question Answering (VQA) have mainly used generic image features from networks trained on the ImageNet dataset, despite the wide scope of questions. In contrast, our approach employs features derived from networks trained for specialized tasks of scene classification, person activ- ity prediction, and person and object attribute prediction. We also present a method for selecting sub-regions of an image that are relevant for evaluating the appropriateness of a putative answer. Visual features are computed both from the whole image and from local regions, while sentences are mapped to a common space using a simple normalized canonical correlation analysis (CCA) model. Our results show a significant improvement over the previous state of the art, and indicate that answering different question types ben- efits from examining a variety of image cues and carefully choosing informative image sub-regions.

Session

Posters 2

Files

Extended Abstract (PDF, 697K)
Paper (PDF, 2M)

DOI

10.5244/C.30.77
https://dx.doi.org/10.5244/C.30.77

Citation

Bibtex

        @inproceedings{BMVC2016_77,
        	title={Solving VIsual Madlibs with Multiple Cues},
        	author={Tatiana Tommasi, Arun Mallya, Bryan Plummer, Svetlana Lazebnik, Alexander Berg and Tamara Berg},
        	year={2016},
        	month={September},
        	pages={77.1-77.13},
        	articleno={77},
        	numpages={13},
        	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
        	publisher={BMVA Press},
        	editor={Richard C. Wilson, Edwin R. Hancock and William A. P. Smith},
        	doi={10.5244/C.30.77},
        	isbn={1-901725-59-6},
        	url={https://dx.doi.org/10.5244/C.30.77}
        }