Summary of Get-up: Geometric-aware Depth Estimation with Radar Points Upsampling, by Huawei Sun et al.
GET-UP: GEomeTric-aware Depth Estimation with Radar Points UPsampling
by Huawei Sun, Zixu Wang, Hao Feng, Julius Ott, Lorenzo Servadei, Robert Wille
First submitted to arxiv on: 2 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
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 GET-UP approach leverages attention-enhanced Graph Neural Networks (GNN) to exchange and aggregate both 2D and 3D information from radar data, effectively enriching the feature representation by incorporating spatial relationships. This method is designed for radar-camera depth estimation in autonomous driving scenarios, where robustness to adverse weather conditions and distance measurements are crucial. GET-UP incorporates a point cloud upsampling task to densify the radar point cloud, rectify point positions, and derive additional 3D features under the guidance of lidar data. The approach fuses radar and camera features during the decoding phase for depth estimation. This paper achieves state-of-the-art performance on the nuScenes dataset with a 15.3% and 14.7% improvement in MAE and RMSE over the previously best-performing model. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary GET-UP is a new way to get better depth estimates from radar data, which is important for self-driving cars. Radar can be used even when it’s raining or dark, making it a reliable source of information. The problem with current methods is that they don’t use all the information in the radar data. GET-UP solves this by using special neural networks to combine 2D and 3D information from the radar data. This helps to create a more accurate picture of what’s around the car. The team also developed a way to make the radar data more detailed, which helps with the accuracy of the depth estimates. |
Keywords
» Artificial intelligence » Attention » Depth estimation » Gnn » Mae
