Summary of Taxonomy-aware Continual Semantic Segmentation in Hyperbolic Spaces For Open-world Perception, by Julia Hindel et al.
Taxonomy-Aware Continual Semantic Segmentation in Hyperbolic Spaces for Open-World Perception
by Julia Hindel, Daniele Cattaneo, Abhinav Valada
First submitted to arxiv on: 25 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Robotics (cs.RO)
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 The proposed Taxonomy-Oriented Poincaré-regularized Incremental-Class Segmentation (TOPICS) model tackles the challenges of class-incremental semantic segmentation by incorporating taxonomy-tree structures into its hyperbolic space-based feature embeddings. This allows for plasticity in old classes, enabling them to adapt to new incremental classes while maintaining a robust structure that prevents catastrophic forgetting. The model is evaluated on eight realistic incremental learning protocols for autonomous driving scenarios, achieving state-of-the-art performance on the Cityscapes and Mapillary Vistas 2.0 benchmarks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary TOPICS is a new way to train semantic segmentation models so they can learn from new classes without forgetting old ones. This is important because real-world data often includes new or unknown objects that need to be recognized. The model uses a special type of space called the Poincaré ball, which helps it remember relationships between different classes. It’s tested on two big datasets and does better than other methods. |
Keywords
» Artificial intelligence » Semantic segmentation
