Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Performance Evaluation of Deep Learning for Tree Branch Segmentation in Autonomous Forestry Systems

Yida Lin, Bing Xue, Mengjie Zhang, Sam Schofield, Richard Green

TL;DR

This work tackles the challenge of tree branch segmentation for UAV-based autonomous forestry by benchmarking 22 configurations across three input resolutions on the Urban Street Tree Dataset. It introduces specialized metrics—Thin Structure IoU (TS-IoU) and Connectivity Preservation Rate (CPR)—to capture thin-branch preservation and topology, complementing IoU and Dice. Key findings show U-Net with MiT-B4 performs best on IoU/Dice at lower resolutions, while MiT-based transformers scale well across resolutions; PSPNet offers the best efficiency with notable accuracy trade-offs. The study provides practical guidance for embedded deployment, highlighting resolution- and backbone-dependent trade-offs and outlining directions for topology-aware loss functions and real-world validation.

Abstract

UAV-based autonomous forestry operations require rapid and precise tree branch segmentation for safe navigation and automated pruning across varying pixel resolutions and operational conditions. We evaluate different deep learning methods at three resolutions (256x256, 512x512, 1024x1024) using the Urban Street Tree Dataset, employing standard metrics (IoU, Dice) and specialized measures including Thin Structure IoU (TS-IoU) and Connectivity Preservation Rate (CPR). Among 22 configurations tested, U-Net with MiT-B4 backbone achieves strong performance at 256x256. At 512x512, MiT-B4 leads in IoU, Dice, TS-IoU, and Boundary-F1. At 1024x1024, U-Net+MiT-B3 shows the best validation performance for IoU/Dice and precision, while U-Net++ excels in boundary quality. PSPNet provides the most efficient option (2.36/9.43/37.74 GFLOPs) with 25.7/19.6/11.8 percentage point IoU reductions compared to top performers at respective resolutions. These results establish multi-resolution benchmarks for accuracy-efficiency trade-offs in embedded forestry systems. Implementation is available at https://github.com/BennyLinntu/PerformanceTreeBranchSegmentation.

Performance Evaluation of Deep Learning for Tree Branch Segmentation in Autonomous Forestry Systems

TL;DR

This work tackles the challenge of tree branch segmentation for UAV-based autonomous forestry by benchmarking 22 configurations across three input resolutions on the Urban Street Tree Dataset. It introduces specialized metrics—Thin Structure IoU (TS-IoU) and Connectivity Preservation Rate (CPR)—to capture thin-branch preservation and topology, complementing IoU and Dice. Key findings show U-Net with MiT-B4 performs best on IoU/Dice at lower resolutions, while MiT-based transformers scale well across resolutions; PSPNet offers the best efficiency with notable accuracy trade-offs. The study provides practical guidance for embedded deployment, highlighting resolution- and backbone-dependent trade-offs and outlining directions for topology-aware loss functions and real-world validation.

Abstract

UAV-based autonomous forestry operations require rapid and precise tree branch segmentation for safe navigation and automated pruning across varying pixel resolutions and operational conditions. We evaluate different deep learning methods at three resolutions (256x256, 512x512, 1024x1024) using the Urban Street Tree Dataset, employing standard metrics (IoU, Dice) and specialized measures including Thin Structure IoU (TS-IoU) and Connectivity Preservation Rate (CPR). Among 22 configurations tested, U-Net with MiT-B4 backbone achieves strong performance at 256x256. At 512x512, MiT-B4 leads in IoU, Dice, TS-IoU, and Boundary-F1. At 1024x1024, U-Net+MiT-B3 shows the best validation performance for IoU/Dice and precision, while U-Net++ excels in boundary quality. PSPNet provides the most efficient option (2.36/9.43/37.74 GFLOPs) with 25.7/19.6/11.8 percentage point IoU reductions compared to top performers at respective resolutions. These results establish multi-resolution benchmarks for accuracy-efficiency trade-offs in embedded forestry systems. Implementation is available at https://github.com/BennyLinntu/PerformanceTreeBranchSegmentation.

Paper Structure

This paper contains 26 sections, 6 equations, 1 figure, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Deep Learning for Semantic Segmentation
  4. Thin Structure Segmentation
  5. Computer Vision in Forestry Applications
  6. Methodology
  7. Problem Formulation
  8. Dataset
  9. Methods Implementation
  10. Evaluation Metrics
  11. Auxiliary Metrics
  12. Evaluation Framework Rationale
  13. Experimental Results
  14. Overall Performance across Resolutions
  15. Performance Analysis
  16. ...and 11 more sections

Figures (1)

  • Figure 1: Example RGB images and ground-truth branch masks from the Urban Street Tree Dataset kendric2023tree.