Toytiny/CMFlow
[CVPR 2023 Highlight 💡] Hidden Gems: 4D Radar Scene Flow Learning Using Cross-Modal Supervision
This project helps autonomous driving engineers precisely understand how objects in a scene are moving, even in challenging weather conditions where cameras might struggle. It takes in 4D radar point cloud data from two consecutive moments and outputs detailed scene flow, showing the velocity of each point, along with motion segmentation and ego-motion estimation. Autonomous vehicle developers and researchers focused on robust perception systems would use this to enhance vehicle awareness.
141 stars. No commits in the last 6 months.
Use this if you need highly accurate, per-point motion estimation from radar data for autonomous vehicles, especially when relying on traditional camera or lidar solutions is problematic.
Not ideal if your application doesn't involve 4D radar data or if you primarily need object detection and tracking rather than detailed scene flow.
Stars
141
Forks
13
Language
Python
License
MIT
Category
Last pushed
Jul 17, 2023
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/computer-vision/Toytiny/CMFlow"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
3DOM-FBK/deep-image-matching
Multiview matching with deep-learning and hand-crafted local features for COLMAP and other SfM...
suhangpro/mvcnn
Multi-view CNN (MVCNN) for shape recognition
zouchuhang/LayoutNet
Torch implementation of our CVPR 18 paper: "LayoutNet: Reconstructing the 3D Room Layout from a...
andyzeng/tsdf-fusion-python
Python code to fuse multiple RGB-D images into a TSDF voxel volume.
andyzeng/tsdf-fusion
Fuse multiple depth frames into a TSDF voxel volume.