Summary of Leveraging Swin Transformer For Local-to-global Weakly Supervised Semantic Segmentation, by Rozhan Ahmadi et al.
Leveraging Swin Transformer for Local-to-Global Weakly Supervised Semantic Segmentation
by Rozhan Ahmadi, Shohreh Kasaei
First submitted to arxiv on: 31 Jan 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 proposes a novel approach to weakly supervised semantic segmentation using image-level labels as supervision. The authors leverage the power of transformers, specifically Swin Transformer (SwinT), to enhance the accuracy of initial seed CAMs (Class Activation Maps). Two variants of the proposed method, SWTformer-V1 and SWTformer-V2, are designed to bring together local and global views. The first variant uses patch tokens as features to generate class probabilities and CAMs, while the second incorporates a multi-scale feature fusion mechanism and a background-aware mechanism for improved localization accuracy. Experimental results on the PascalVOC 2012 dataset show that both variants outperform state-of-the-art models, with SWTformer-V1 achieving a 0.98% mAP higher localization accuracy and SWTformer-V2 improving the accuracy of generated seed CAMs by 5.32% mIoU. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps computer vision get better at understanding images without needing lots of labeled data. They use special kind of AI model called Swin Transformer to make predictions about what’s in an image. The authors create two new methods, SWTformer-V1 and SWTformer-V2, that work together to make the predictions more accurate. They test these methods on a big dataset and find that they do better than other ways of doing this kind of prediction. |
Keywords
» Artificial intelligence » Semantic segmentation » Supervised » Transformer
