Summary of Shelf-supervised Cross-modal Pre-training For 3d Object Detection, by Mehar Khurana et al.
Shelf-Supervised Cross-Modal Pre-Training for 3D Object Detection
by Mehar Khurana, Neehar Peri, James Hays, Deva Ramanan
First submitted to arxiv on: 14 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: 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 A novel self-supervised pre-training method for 3D object detectors is proposed, leveraging image-based foundation models trained on internet-scale data to generate pseudo-labels for LiDAR point clouds. This approach, dubbed shelf-supervision, enables the training of zero-shot 3D bounding boxes from paired RGB and LiDAR data. The method is shown to significantly improve semi-supervised detection accuracy compared to prior self-supervised pretext tasks on nuScenes and WOD datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new way to train 3D object detectors without labeling every point cloud is developed. Instead of using only the 3D data, it combines this with images from the same scene, allowing for more effective training. This approach uses pre-trained image models to generate fake labels for the 3D data, making it possible to learn from much smaller and less diverse datasets. |
Keywords
» Artificial intelligence » Self supervised » Semi supervised » Zero shot
