arXiv Computer Vision August 28, 2026 By Trung Thanh Nguyen, Daniel Lusk, Kilian Gerberding, Janusch Vajna-Jehle, Tuan-Anh Vu, Duc Viet Le, Tu Vo, Phi Le Nguyen, Yasutomo Kawanishi, Takahiro Komamizu, Ichiro Ide, Julian Frey, Teja Kattenborn

SelectAnyTree: A Promptable Instance Segmentation Model for 3D Forest LiDAR Point Clouds

Read the original on arXiv Computer Vision →

SelectAnyTree is a promptable instance segmentation model designed for 3D forest LiDAR point clouds, enabling users to delineate individual trees with a few clicks. The architecture comprises a sparse voxel scene encoder, a click‑to‑query prompt encoder, and a state‑space query decoder that produces tree masks in linear time, requiring only 19.4 M parameters. Across seven forest regions and an independent dataset, the model achieves a 79.9 % IoU for a single‑click target tree, outperforming existing promptable baselines and requiring the fewest clicks to reach accuracy targets.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computer Vision.

computer-vision

arXiv Machine Learning

Jun 9

SegmentAnyTreeV2: Scaling Transformer-Based Tree Instance Segmentation Across Sensors, Platforms, and Forests

arXiv:2606. 08206v1 Announce Type: cross Abstract: We present SegmentAnyTreeV2, a sensor- and platform-agnostic framework for semantic and instance segmentation of forest point clouds.

By Maciej Wielgosz, Stefano Puliti, Rasmus Astrup

llms computer-vision benchmarks

arXiv Computer Vision

Sep 1

CedarCypress3D: an annotated UAV-LiDAR dataset of individual trees in planted cedar and cypress forests

arXiv:2608.30149v1 Announce Type: new Abstract: Individual tree measurements derived from Light Detection and Ranging (LiDAR) mounted on Unmanned Aerial Vehicles (UAV) provide valuable information fo...

By Katsuto Shimizu (Shikoku Research Center, Forestry and Forest Products Research Institute), Fumiaki Kitahara (Department of Forest Management, Forestry and Forest Products Research Institute), Tomohiro Nishizono (Department of Forest Management, Forestry and Forest Products Research Institute), Hideki Saito (Forestry and Forest Products Research Institute), Masayoshi Takahashi (Department of Forest Management, Forestry and Forest Products Research Institute), Shingo Obata (Hokkaido Research Center, Forestry and Forest Products Research Institute), Shunsuke Tei (Hokkaido Research Center, Forestry and Forest Products Research Institute), Naoyuki Furuya (Department of Forest Management, Forestry and Forest Products Research Institute), Tomoya Goto (Green Kogyo), Eiji Kodani (Department of Forest Management, Forestry and Forest Products Research Institute), Yusuke Yamada (Graduate School of Bioagricultural Sciences, Nagoya University)

computer-vision

arXiv AI

Jul 9

Self-Supervised Pretraining Improves Cross-Site and Cross-Scale Robustness of Point Cloud Leaf-Wood Segmentation

arXiv:2607. 06948v1 Announce Type: cross Abstract: The accuracy of existing leaf-wood segmentation methods for tree point clouds varies across forest types and sites.

By Heeju Mun, Tackang Yang, Yunsoo Nam, Changhyun Choi

computer-vision fine-tuning benchmarks

arXiv Computer Vision

Aug 31

Bringing SAM to new heights: Leveraging elevation data for tree crown segmentation from drone imagery

The paper introduces BalSAM, a model that combines the Segment Anything Model (SAM) with Digital Surface Model (DSM) elevation data to improve tree crown instance segmentation from high‑resolution drone imagery. Experiments across boreal plantations, temperate forests, and tropical forests show that while off‑the‑shelf SAM does not beat a custom Mask R-CNN, fine‑tuning SAM end‑to‑end and incorporating DSM information yield promising results, especially for plantation sites.

By M\'elisande Teng, Arthur Ouaknine, Etienne Lalibert\'e, Yoshua Bengio, David Rolnick, Hugo Larochelle

computer-vision robotics

arXiv AI

Jul 8

NegROI: Click-Centric Uncertainty-Guided Refinement with Scene-Conditioned Negative Prompts for Robust Interactive 3D Segmentation

arXiv:2607. 05955v1 Announce Type: cross Abstract: Interactive 3D segmentation aims to extract object masks in point clouds with minimal user clicks.

By Shuheng Zhang, Feng Wu

llms computer-vision benchmarks

arXiv AI

Jul 1

FLORA: A deep learning approach to predict forest attributes from heterogeneous LiDAR data

arXiv:2606. 32023v1 Announce Type: cross Abstract: Forest attributes are essential for national-scale resource monitoring.

By Emilie Vautier, Cl\'ement Mallet, C\'edric Vega

One email a morning, machine-written

Related stories

SegmentAnyTreeV2: Scaling Transformer-Based Tree Instance Segmentation Across Sensors, Platforms, and Forests

CedarCypress3D: an annotated UAV-LiDAR dataset of individual trees in planted cedar and cypress forests

Self-Supervised Pretraining Improves Cross-Site and Cross-Scale Robustness of Point Cloud Leaf-Wood Segmentation

Bringing SAM to new heights: Leveraging elevation data for tree crown segmentation from drone imagery

NegROI: Click-Centric Uncertainty-Guided Refinement with Scene-Conditioned Negative Prompts for Robust Interactive 3D Segmentation

FLORA: A deep learning approach to predict forest attributes from heterogeneous LiDAR data