Summary of First Place Solution to the Multiple-choice Video Qa Track Of the Second Perception Test Challenge, by Yingzhe Peng et al.
First Place Solution to the Multiple-choice Video QA Track of The Second Perception Test Challenge
by Yingzhe Peng, Yixiao Yuan, Zitian Ao, Huapeng Zhou, Kangqi Wang, Qipeng Zhu, Xu Yang
First submitted to arxiv on: 20 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
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 presents the winning solution for the Multiple-choice Video Question Answering (QA) track in The Second Perception Test Challenge. The task requires models to accurately comprehend and answer questions about video content, which is a complex problem that demands powerful video understanding capabilities. To tackle this challenge, the authors leverage the QwenVL2 (7B) model and fine-tune it on the provided training set, while also employing ensemble strategies and test-time augmentation techniques to boost performance. As a result, their approach achieves a Top-1 Accuracy of 0.7647 on the leaderboard. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about creating a computer program that can understand videos and answer questions about them. This is a difficult task because computers need to be able to recognize what’s happening in the video and then figure out the right answer. The authors use a powerful model called QwenVL2 (7B) and make it better by training it on lots of examples. They also try different ways to improve its performance, like combining multiple models together and trying different versions of the same question. By doing this, they were able to create a program that can answer questions about videos very accurately. |
Keywords
» Artificial intelligence » Question answering
