Summary of Object Attribute Matters in Visual Question Answering, by Peize Li et al.
Object Attribute Matters in Visual Question Answering
by Peize Li, Qingyi Si, Peng Fu, Zheng Lin, Yan Wang
First submitted to arxiv on: 20 Dec 2023
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel approach to visual question answering (VQA) by leveraging object attributes as a bridge to unify visual and textual information. The proposed method incorporates an attribute fusion module that fuses attributes and visual features using message passing, leading to better object-level visual-language alignment. This is achieved through a multimodal graph neural network that constructs a graph of objects and their attributes. The paper also introduces a contrastive knowledge distillation module that distills implicit knowledge from the model’s representation learning, further enhancing the scene understanding. Experimental results on six datasets, including COCO-QA, VQAv2, and TDIUC, demonstrate the superiority of the proposed method. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine trying to answer a question about a picture by reading its caption. That’s what visual question answering is all about! But it’s hard because we need to understand both the picture and the words at the same time. This paper has an idea for how to make that easier. They propose using object attributes, like “is this a cat or a dog?” to help match up the picture and text. This makes it better at understanding what’s happening in the scene. The authors also test their method on six different datasets and show that it does better than other approaches. |
Keywords
» Artificial intelligence » Alignment » Graph neural network » Knowledge distillation » Question answering » Representation learning » Scene understanding
