Question Type Guided Attention in Visual Question Answering
Abstract
Visual Question Answering (VQA) requires integration of feature maps with drastically different structures. Image descriptors have structures at multiple spatial scales, while lexical inputs inherently follow a temporal sequence and naturally cluster into semantically different question types. A lot of previous works use complex models to extract feature representations but neglect to use high-level information summary such as question types in learning. In this work, we propose Question Type-guided Attention (QTA). It utilizes the information of question type to dynamically balance between bottom-up and top-down visual features, respectively extracted from ResNet and Faster R-CNN networks. We experiment with multiple VQA architectures with extensive input ablation studies over the TDIUC dataset and show that QTA systematically improves the performance by more than 5% across multiple question type categories such as "Activity Recognition", "Utility" and "Counting" on TDIUC dataset compared to the state-of-art. By adding QTA on the state-of-art model MCB, we achieve 3% improvement in overall accuracy. Finally, we propose a multi-task extension to predict question types which generalizes QTA to applications that lack question type, with a minimal performance loss.
Additional Information
© Springer Nature Switzerland AG 2018. Work partially done while the author was working at Amazon AI. We thank Amazon AI for providing computing resources. Yang Shi is supported by Air Force Award FA9550-15-1-0221.Attached Files
Accepted Version - 1804.02088.pdf
Accepted Version - Yang_Shi_Question_Type_Guided_ECCV_2018_paper.pdf
Files
Name | Size | Download all |
---|---|---|
md5:6c4d910afe5d348a2f15320bc3363e79
|
2.9 MB | Preview Download |
md5:5ec8ce98250e16fac8058024106722e8
|
1.2 MB | Preview Download |
Additional details
- Eprint ID
- 94175
- Resolver ID
- CaltechAUTHORS:20190327-085753056
- Amazon AI
- FA9550-15-1-0221
- Air Force Office of Scientific Research (AFOSR)
- Created
-
2019-03-29Created from EPrint's datestamp field
- Updated
-
2021-11-16Created from EPrint's last_modified field
- Series Name
- Lecture Notes in Computer Science
- Series Volume or Issue Number
- 11208