Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

In Defense of Grid Features for Visual Question Answering

About

Popularized as 'bottom-up' attention, bounding box (or region) based visual features have recently surpassed vanilla grid-based convolutional features as the de facto standard for vision and language tasks like visual question answering (VQA). However, it is not clear whether the advantages of regions (e.g. better localization) are the key reasons for the success of bottom-up attention. In this paper, we revisit grid features for VQA, and find they can work surprisingly well - running more than an order of magnitude faster with the same accuracy (e.g. if pre-trained in a similar fashion). Through extensive experiments, we verify that this observation holds true across different VQA models (reporting a state-of-the-art accuracy on VQA 2.0 test-std, 72.71), datasets, and generalizes well to other tasks like image captioning. As grid features make the model design and training process much simpler, this enables us to train them end-to-end and also use a more flexible network design. We learn VQA models end-to-end, from pixels directly to answers, and show that strong performance is achievable without using any region annotations in pre-training. We hope our findings help further improve the scientific understanding and the practical application of VQA. Code and features will be made available.

Huaizu Jiang, Ishan Misra, Marcus Rohrbach, Erik Learned-Miller, Xinlei Chen• 2020

Related benchmarks

TaskDatasetResultRank
Visual Question AnsweringVizWiz
Accuracy54.17
1043
Image CaptioningMS COCO Karpathy (test)
CIDEr1.138
682
Visual Question AnsweringVQA v2 (test-dev)
Overall Accuracy72.59
664
Visual Question AnsweringVQA v2 (test-std)--
466
Visual Question AnsweringVQA 2.0 (test-dev)
Accuracy72.59
337
Attribute PredictionCityscapes Attributes Recognition (CAR) 44
mA66.8
5
Attribute PredictionVisual Attributes in the Wild (VAW) 53
Mean Accuracy50.3
5
Showing 7 of 7 rows

Other info

Follow for update