Summary of Early Preparation Pays Off: New Classifier Pre-tuning For Class Incremental Semantic Segmentation, by Zhengyuan Xie et al.
Early Preparation Pays Off: New Classifier Pre-tuning for Class Incremental Semantic Segmentation
by Zhengyuan Xie, Haiquan Lu, Jia-wen Xiao, Enguang Wang, Le Zhang, Xialei Liu
First submitted to arxiv on: 19 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: 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 proposes a novel approach to address the issue of catastrophic forgetting and background shift in class incremental semantic segmentation, where new classifiers are initialized for learning new tasks while preserving old knowledge. The method, called NeST (Neural Segmenter Transformation), learns a transformation from old classifiers to generate new classifiers rather than directly tuning their parameters. This allows new classifiers to align with the backbone and adapt to new data, preventing drastic changes in the feature extractor when learning new classes. To achieve this stability-plasticity trade-off, the authors also design a strategy considering cross-task class similarity for initializing matrices used in the transformation. Experimental results on Pascal VOC 2012 and ADE20K datasets show that NeST can significantly improve the performance of previous methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us learn new things without forgetting what we already know. It’s like when you’re learning a new language, but you don’t want to forget how to speak your old language too! The authors came up with a clever way to teach computers to do this by creating a “map” from the old way of thinking to the new way. This helps the computer learn new things without getting confused or forgetting what it already knows. They tested their idea on some pictures and found that it works really well! |
Keywords
* Artificial intelligence * Semantic segmentation
