HAAG LiDAR, Georgia Tech

We work on terrestrial LiDAR scans of vegetation: separating individual trees in a scan and reconstructing their branch structure, so habitat can be measured tree by tree. The work is part of the CS 6999 HAAG LiDAR Terrestrial Vegetation Analysis project at Georgia Tech.

Models

• point-sam-wytham-lora: Point-SAM fine-tuned to outline individual trees from a click, trained on the ISPRS Wytham benchmark. A research checkpoint; its model card gives the results and limits.

Code

• Training and evaluation: https://github.com/nisulay/pointsam-training
• Team repositories: https://github.com/Landscape-CV

Contact

Open a discussion on a model page, or an issue on the matching GitHub repository.