Summary of The Solution For Temporal Action Localisation Task Of Perception Test Challenge 2024, by Yinan Han et al.
The Solution for Temporal Action Localisation Task of Perception Test Challenge 2024
by Yinan Han, Qingyuan Jiang, Hongming Mei, Yang Yang, Jinhui Tang
First submitted to arxiv on: 8 Oct 2024
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 presents a method for Temporal Action Localisation (TAL), which aims to identify and classify actions within specific time intervals in video sequences. To enhance generalizability, the authors employ data augmentation by expanding the training dataset using overlapping labels from the Something-SomethingV2 dataset. State-of-the-art models are used for feature extraction, including UMT, VideoMAEv2, BEATs, and CAV-MAE for video and audio features. The approach involves training multimodal and unimodal models separately, then combining their predictions using Weighted Box Fusion (WBF). This fusion strategy ensures robust action localisation. The overall approach achieves a score of 0.5498, securing first place in the competition. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us better understand how to find specific actions within videos. It’s like trying to find a specific moment in a movie where something important happens. To do this, the researchers used special models that can learn from lots of video examples and combine what they see with what they hear (like music or sound effects). They tested their method on a big competition and came out on top! |
Keywords
» Artificial intelligence » Data augmentation » Feature extraction » Mae
