Summary of Odin: a Single Model For 2d and 3d Segmentation, by Ayush Jain et al.
ODIN: A Single Model for 2D and 3D Segmentation
by Ayush Jain, Pushkal Katara, Nikolaos Gkanatsios, Adam W. Harley, Gabriel Sarch, Kriti Aggarwal, Vishrav Chaudhary, Katerina Fragkiadaki
First submitted to arxiv on: 4 Jan 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); 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 This paper challenges the notion that 2D and 3D perception require distinct model architectures. It proposes a transformer-based model called ODIN, which can segment and label both 2D RGB images and 3D point clouds by alternately fusing 2D within-view and 3D cross-view information. The model achieves state-of-the-art performance on several instance segmentation benchmarks, including ScanNet200, Matterport3D, AI2THOR 3D, and competitive performance on others. ODIN outperforms previous works when using sensed 3D point clouds or as a 3D perception engine in an instructable embodied agent architecture. |
Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper shows that one model can do both 2D and 3D tasks. It makes a special computer program called ODIN that can look at pictures and 3D maps and find things like chairs, tables, and other objects. This is useful because it means we don’t need to make separate programs for each task. The program does really well on tests and even helps robots understand what they’re seeing. |
Keywords
* Artificial intelligence * Instance segmentation * Transformer