Publications

Seven peer-reviewed papers, first author on all of them.

Full list on Google Scholar

Peer-reviewed

Springer LNCS 2026
Progressive Per-Branch Depth Optimization with DEFOM-Stereo and SAM3 for UAV Forestry
Turns noisy disparity into clean per-branch point clouds. A five-stage filter cuts depth deviation by 82%.

Springer arXiv
Springer LNCS 2026
UE5-Forest: Photorealistic Synthetic Stereo Pairs with Pixel-Exact Disparity for UAV Forestry Depth Estimation
Builds a synthetic stereo dataset in Unreal Engine 5: 115 scanned trees and 5,520 pairs with pixel-exact labels.

Proceedings arXiv
Springer LNCS 2026
From Pseudo-Ground-Truth to Edge Deployment: Benchmarking Deep Stereo Matching Networks on Real Radiata Pine Tree Images for UAV Tree Pruning
Trains ten deep stereo networks on 5,313 real pine stereo pairs, then times them on the drone's Jetson.

Proceedings arXiv
IEEE IVCNZ 2025
YOLO and SGBM Integration for Autonomous Tree Branch Detection and Depth Estimation in Radiata Pine Pruning Applications
Detects and positions branches from stereo alone, with no LiDAR: 82.0% mask mAP_50-95 across a 2 m working range.

IEEE arXiv
IEEE IVCNZ 2025
Performance Evaluation of Deep Learning for Tree Branch Segmentation in Autonomous Forestry Systems
Compares 22 segmentation configurations at three resolutions with thin-structure metrics, charting accuracy against compute for embedded use.

IEEE arXiv Code
IEEE IVCNZ 2025
Genetic Algorithms for Parameter Optimization for Disparity Map Generation of Radiata Pine Branch Images
Replaces manual SGBM and WLS tuning with a genetic algorithm: 42.86% lower MSE and 28.52% higher SSIM.

IEEE arXiv
IEEE IVCNZ 2024
Deep Learning-Based Depth Map Generation and YOLO-Integrated Distance Estimation for Radiata Pine Branch Detection Using Drone Stereo Vision
Shows deep depth estimation beats SGBM on branches, and combines it with YOLO masks to measure branch distance.

IEEE arXiv

Recent preprints

arXiv 2026
Real-Time Branch-to-Tool Distance Estimation for Autonomous UAV Pruning
Five DEFOM-Stereo variants, from simulation to Jetson deployment.
arXiv 2026
EMCStereo: Attention-Enhanced Stereo Matching for Thin-Structure Depth Estimation
A lighter attention stack for thin structures: 1.31 px EPE on VirtualTree and 0.62 px on ETH3D.
arXiv 2026
Low-Cost Stereo Vision for Robust 3D Positioning of Thin Radiata Pine Branches
Centroid triangulation with outlier rejection gives one robust distance per branch, without auxiliary depth sensors.
arXiv 2026
Towards Gold-Standard Depth Estimation for Tree Branches in UAV Forestry
Zero-shot benchmark of eight stereo methods that sets DEFOM as the reference for vegetation.
arXiv 2026
PACT: Pairwise-Anchored Calibrated Tuning for Single-Token Typed Decisions
Sole author. Calibrated fine-tuning for LLMs that answer in one token.