Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automatic Pith Detection in Tree Cross-Section Images Using Deep Learning

Tzu-I Liao, Mahmoud Fakhry, Jibin Yesudas Varghese

TL;DR

This work tackles automated pith detection in tree cross-sections using five deep learning models. It conducts a comprehensive evaluation on a 582-image dataset with dynamic augmentation and a post-processing enhancement (NMS) for Mask R-CNN, plus a generalization test on an unseen oak dataset. The Swin Transformer achieves the best localization accuracy and segmentation quality, while YOLOv9 provides fast inference with coarser boundaries; DeepLabV3 and U-Net offer balanced trade-offs, and Mask R-CNN benefits greatly from NMS. The findings guide model selection based on dataset characteristics and deployment constraints, and highlight the importance of post-processing and data diversity for robust forestry applications.

Abstract

Pith detection in tree cross-sections is essential for forestry and wood quality analysis but remains a manual, error-prone task. This study evaluates deep learning models -- YOLOv9, U-Net, Swin Transformer, DeepLabV3, and Mask R-CNN -- to automate the process efficiently. A dataset of 582 labeled images was dynamically augmented to improve generalization. Swin Transformer achieved the highest accuracy (0.94), excelling in fine segmentation. YOLOv9 performed well for bounding box detection but struggled with boundary precision. U-Net was effective for structured patterns, while DeepLabV3 captured multi-scale features with slight boundary imprecision. Mask R-CNN initially underperformed due to overlapping detections, but applying Non-Maximum Suppression (NMS) improved its IoU from 0.45 to 0.80. Generalizability was next tested using an oak dataset of 11 images from Oregon State University's Tree Ring Lab. Additionally, for exploratory analysis purposes, an additional dataset of 64 labeled tree cross-sections was used to train the worst-performing model to see if this would improve its performance generalizing to the unseen oak dataset. Key challenges included tensor mismatches and boundary inconsistencies, addressed through hyperparameter tuning and augmentation. Our results highlight deep learning's potential for tree cross-section pith detection, with model choice depending on dataset characteristics and application needs.

Automatic Pith Detection in Tree Cross-Section Images Using Deep Learning

TL;DR

This work tackles automated pith detection in tree cross-sections using five deep learning models. It conducts a comprehensive evaluation on a 582-image dataset with dynamic augmentation and a post-processing enhancement (NMS) for Mask R-CNN, plus a generalization test on an unseen oak dataset. The Swin Transformer achieves the best localization accuracy and segmentation quality, while YOLOv9 provides fast inference with coarser boundaries; DeepLabV3 and U-Net offer balanced trade-offs, and Mask R-CNN benefits greatly from NMS. The findings guide model selection based on dataset characteristics and deployment constraints, and highlight the importance of post-processing and data diversity for robust forestry applications.

Abstract

Pith detection in tree cross-sections is essential for forestry and wood quality analysis but remains a manual, error-prone task. This study evaluates deep learning models -- YOLOv9, U-Net, Swin Transformer, DeepLabV3, and Mask R-CNN -- to automate the process efficiently. A dataset of 582 labeled images was dynamically augmented to improve generalization. Swin Transformer achieved the highest accuracy (0.94), excelling in fine segmentation. YOLOv9 performed well for bounding box detection but struggled with boundary precision. U-Net was effective for structured patterns, while DeepLabV3 captured multi-scale features with slight boundary imprecision. Mask R-CNN initially underperformed due to overlapping detections, but applying Non-Maximum Suppression (NMS) improved its IoU from 0.45 to 0.80. Generalizability was next tested using an oak dataset of 11 images from Oregon State University's Tree Ring Lab. Additionally, for exploratory analysis purposes, an additional dataset of 64 labeled tree cross-sections was used to train the worst-performing model to see if this would improve its performance generalizing to the unseen oak dataset. Key challenges included tensor mismatches and boundary inconsistencies, addressed through hyperparameter tuning and augmentation. Our results highlight deep learning's potential for tree cross-section pith detection, with model choice depending on dataset characteristics and application needs.

Paper Structure

This paper contains 23 sections, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Motivation
  4. Contributions
  5. Literature Review
  6. Traditional Methods
  7. Deep Learning Approaches
  8. Comparison of Traditional vs. Deep Learning Approaches
  9. Methodology
  10. Dataset
  11. Model Architectures and Training
  12. Swin Transformer
  13. YOLOv9
  14. DeepLabV3
  15. U-Net
  16. ...and 8 more sections

Figures (7)

  • Figure 1: Example of an input tree cross-section image from the dataset.
  • Figure 2: Oak dataset example tree cross-section for pith detection (top) and an example tree cross-section pith detection from the 582 images (bottom). The red box denotes the predicted bounding box while the green box denotes the ground truth bounding box.
  • Figure 3: Training and Validation Loss Curve for Swin Transformer.
  • Figure 4: Training and Validation Loss Curve for YOLOv9.
  • Figure 5: Training and Validation Loss Curve for DeepLabV3.
  • ...and 2 more figures