Summary of Efficient Selective Audio Masked Multimodal Bottleneck Transformer For Audio-video Classification, by Wentao Zhu
Efficient Selective Audio Masked Multimodal Bottleneck Transformer for Audio-Video Classification
by Wentao Zhu
First submitted to arxiv on: 8 Jan 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multimedia (cs.MM); Sound (cs.SD); Audio and Speech Processing (eess.AS)
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 A novel audio-video recognition approach, dubbed Audio Video Transformer (AVT), is proposed to effectively learn from multimodal videos. By leveraging video Transformer’s spatio-temporal representation and reducing cross-modality complexity through an audio-video bottleneck Transformer, AVT improves action recognition accuracy. To boost learning efficiency, self-supervised objectives such as contrastive learning, masked audio and video learning, and audio-video matching are integrated into AVT training. A masked audio segment loss is also introduced to learn semantic audio activities. Experimental results on three public datasets and two in-house datasets demonstrate the effectiveness of AVT, outperforming previous state-of-the-art counterparts by 8% on Kinetics-Sounds and surpassing a previous state-of-the-art video Transformer by 10% on VGGSound. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers has developed a new way to understand videos that include both audio and visuals. They created an “Audio Video Transformer” (AVT) that can learn from these multimodal videos more effectively than before. This is done by using the video’s spatial and temporal information, reducing the complexity between audio and video, and making the learning process more efficient. The AVT was tested on several datasets and showed improved performance compared to previous methods. |
Keywords
* Artificial intelligence * Self supervised * Transformer
