A new unified framework for supervised 3D crown segmentation (TreeisoNet) using deep neural networks across airborne, UAV-borne, and terrestrial laser scans
Bibliographic record
Abstract
Accurately defining and isolating 3D tree space is critical for extracting and analyzing tree inventory attributes, yet it remains a challenge due to the structural complexity and heterogeneity within natural forests. This study introduces TreeisoNet, a suite of supervised deep neural networks tailored for robust 3D tree segmentation across natural forest environments. These networks are specifically designed to identify tree locations, stem components (if available), and crown clusters, making them adaptable to varying scales of laser scanning from airborne laser scannner (ALS), terrestrial laser scanner (TLS), and unmanned aerial vehicle (UAV). Our evaluation used three benchmark datasets with manually isolated tree references, achieving mean intersection-over-union (mIoU) accuracies of 0.81 for UAV, 0.76 for TLS, and 0.59 for ALS, which are competitive with contemporary algorithms such as ForAINet, Treeiso, Mask R-CNN, and AMS3D. Noise from stem point delineation minimally impacted stem location detection but significantly affected crown clustering. Moderate manual refinement of stem points or tree centers significantly improved tree segmentation accuracies, achieving 0.85 for UAV, 0.86 for TLS, and 0.80 for ALS. The study confirms SegFormer as an effective 3D point-level classifier and an offset-based UNet as a superior segmenter, with the latter outperforming unsupervised solutions like watershed and shortest-path methods. TreeisoNet demonstrates strong adaptability in capturing invariant tree geometry features, ensuring transferability across different resolutions, sites, and sensors with minimal accuracy loss.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".