Summary of Segment, Lift and Fit: Automatic 3d Shape Labeling From 2d Prompts, by Jianhao Li et al.
Segment, Lift and Fit: Automatic 3D Shape Labeling from 2D Prompts
by Jianhao Li, Tianyu Sun, Zhongdao Wang, Enze Xie, Bailan Feng, Hongbo Zhang, Ze Yuan, Ke Xu, Jiaheng Liu, Ping Luo
First submitted to arxiv on: 16 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 This paper proposes an algorithm called Segment, Lift, and Fit (SLF) that automatically labels 3D objects from 2D point or box prompts. Unlike previous approaches, SLF predicts 3D shapes rather than bounding boxes and doesn’t require training on a specific dataset. The method first segments high-quality instance masks using the Segment Anything Model (SAM), then lifts the 2D masks to 3D forms and adjusts their poses and shapes until they fit the original prompts. The authors claim that this approach generalizes well across different datasets, as it doesn’t overfit to biased annotation patterns. Experimental results on the KITTI dataset demonstrate that SLF produces high-quality bounding box annotations (AP@0.5 IoU of nearly 90%) and shows promising results in detailed shape predictions. The paper’s contributions include a novel approach for occupancy annotation of dynamic objects. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us teach computers to recognize 3D objects by giving them simple descriptions. Instead of trying to find the exact shape, we describe what the object looks like from different angles. This is useful for self-driving cars that need to understand the world around them. The authors came up with a new way to do this without needing to train on specific data sets. They tested their method on real car data and it worked really well. |
Keywords
» Artificial intelligence » Bounding box » Sam
