Summary of Multi-object Event Graph Representation Learning For Video Question Answering, by Yanan Wang and Shuichiro Haruta and Donghuo Zeng and Julio Vizcarra and Mori Kurokawa
Multi-object event graph representation learning for Video Question Answering
by Yanan Wang, Shuichiro Haruta, Donghuo Zeng, Julio Vizcarra, Mori Kurokawa
First submitted to arxiv on: 12 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
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 video question answering (VideoQA), which involves predicting the correct answer to questions about videos. The proposed method, called CLanG, aims to capture complex scenarios involving multiple objects by employing a multi-layer GNN-cluster module for adversarial graph representation learning. This allows for contrastive learning between the question text and its relevant multi-object event graph. The authors demonstrate the effectiveness of their approach on two challenging VideoQA datasets, NExT-QA and TGIF-QA-R, achieving up to 2.2% higher accuracy than a strong baseline. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The proposed method, CLanG, addresses the limitation of prior works that focused on modeling individual object movements using transformer-based methods. By capturing event representations associated with multiple objects, CLanG enables reasoning about complex scenarios, such as “a boy is throwing a ball in a hoop”. The authors also highlight the strength of their approach in handling causal and temporal questions, achieving 2.8% better accuracy than baselines. |
Keywords
» Artificial intelligence » Gnn » Question answering » Representation learning » Transformer
