ForestFormer3D: A Unified Framework for End-to-End Segmentation of Forest LiDAR 3D Point Clouds

arXiv Code Dataset
Semantic view Instance view

ForestFormer3D achieves new state‑of‑the‑art individual tree segmentation and scene‑level semantic segmentation in complex forests

Abstract

The segmentation of forest LiDAR 3D point clouds, including both individual tree and semantic segmentation, is fundamental for advancing forest management and ecological research. However, current approaches often struggle with the complexity and variability of natural forest environments.

We present ForestFormer3D, a new unified and end-to-end framework designed for precise individual tree and semantic segmentation. ForestFormer3D incorporates ISA-guided query point selection, a score-based block merging strategy during inference, and a one-to-many association mechanism for effective training. By combining these new components, our model achieves state-of-the-art performance for individual tree segmentation on the newly introduced FOR-instanceV2 dataset, which spans diverse forest types and regions.

Additionally, ForestFormer3D generalizes well to unseen test sets (Wytham woods and LAUTx), showcasing its robustness across different forest conditions and sensor modalities. The FOR-instanceV2 dataset and the ForestFormer3D code are publicly available.

Illustration of our ForestFormer3D framework

Framework Illustration

Geographical overview of the FOR‑instanceV2 dataset

FOR‑instanceV2 overview

Comparison with other methods

Results on the FOR‑instanceV2 test split. The best results are in bold, and the second best ones are underlined.

Method comparison

Qualitative results

GT
Prediction
CULS CULS
NIBIO_MLS NIBIO_MLS
NIBIO plot1 NIBIO plot1
Blue Cat Blue Cat
NIBIO2 plot3 NIBIO2 plot3
RMIT RMIT
SCION plot31 SCION plot31
TUWIEN TUWIEN
Yuchen Yuchen

BibTeX 🙏 

@inproceedings{xiang2025forestformer3d,
  title     = {ForestFormer3D: A Unified Framework for End-to-End Segmentation of Forest LiDAR 3D Point Clouds},
  author    = {Binbin Xiang and Maciej Wielgosz and Stefano Puliti and Kamil Kr{\'a}l and Martin Kr\r{u}\v{c}ek and Azim Missarov and Rasmus Astrup},
  booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  year      = {2025}
}