Summary of Weaksam: Segment Anything Meets Weakly-supervised Instance-level Recognition, by Lianghui Zhu et al.
WeakSAM: Segment Anything Meets Weakly-supervised Instance-level Recognition
by Lianghui Zhu, Junwei Zhou, Yan Liu, Xin Hao, Wenyu Liu, Xinggang Wang
First submitted to arxiv on: 22 Feb 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 introduces WeakSAM, a weakly supervised visual recognition model that leverages the pre-trained Segment Anything Model (SAM) to tackle object detection and segmentation tasks. Unlike traditional approaches, which rely on pseudo-labeling and multi-instance learning, WeakSAM addresses limitations in pseudo ground truth generation and noisy instances through adaptive PGT generation and Region of Interest drop regularization. The proposed method also overcomes the need for prompts and category awareness in SAM-based automatic object detection and segmentation. Experimental results demonstrate significant improvements over previous state-of-the-art methods in weakly-supervised object detection (WSOD) and segmentation (WSIS) benchmarks, with average gains of 7.4% and 8.5%, respectively. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about a new way to teach computers to recognize objects in images without needing humans to label every single one. Instead, it uses a pre-trained model called Segment Anything Model (SAM) to help with object detection and segmentation. The approach addresses some common issues in this type of learning, like incomplete or noisy labels, by generating better pseudo labels and dropping unnecessary information. The results show that this new method performs much better than previous ones on similar tasks, making it a promising step forward for computer vision. |
Keywords
» Artificial intelligence » Object detection » Regularization » Sam » Supervised
