A new article has just bin published in Remote Sensing of Environment featuring SegmentAnyTree: A sensor platform and agnostic deep learning model for tree segmentation using laser scanning data.
Congratulations to the authors Maciej Wielgosz, Stefano Puliti, Binbin Xiang, Konrad Schindler and Rasmus Astrup on their work!
This study focuses on advancing individual tree crown segmentation in lidar data, developing a sensor- and platform-agnostic deep learning model transferable across a spectrum of dense laser scanning datasets from drone , to terrestrial, and mobile laser scanning data. In a field where transferability across different data characteristics has been a longstanding challenge, this research marks a step towards versatile, efficient, and comprehensive 3D forest scene analysis.
The article highlights are:
- Developed a versatile deep learning method for tree segmentation with lidar.
- Evaluated model performance across various platforms with sparsity impact analysis.
- Showed that gradual sparsification as a training strategy is effective for lidar.
- Improved understory tree detection with lower computational needs than others.
- Set new performance standards in Wytham Woods and TreeLearn datasets.
Read and download the full article here
The FOR-Instance dataset is available here