LeafInst - Unified Instance Segmentation Network for Fine-Grained Forestry Leaf Phenotype Analysis: A New UAV based Benchmark

2026-03-03

AI Summary

This study addresses the lack of fine-grained instance segmentation methods for young tree leaves in natural forest environments, a domain underexplored compared to large-leaf crops and challenged by significant scale variation, complex lighting, and irregular leaf morphology. To bridge this gap, the authors introduce Poplar-leaf, the first UAV-based image dataset dedicated to wild tree leaves, and propose LeafInst, a unified instance segmentation network. LeafInst integrates an Asymptotic Feature Pyramid Network (AFPN), a Dynamic Asymmetric Spatial Perception (DASP) module, and a Dual Residual Anomaly Regression Head with Top-down Cross-scale Feature Upsampling (DARH-TCFU) to effectively model multi-scale and irregular leaf structures. Experiments demonstrate that LeafInst achieves 68.4 mAP on Poplar-leaf, outperforming YOLOv11 and MaskDINO, and attains 52.7 box mAP on PhenoBench, highlighting its strong generalization and practical utility.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Segmentation
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Intelligent forest tree breeding has advanced plant phenotyping, yet existing research largely focuses on large-leaf agricultural crops, with limited attention to fine-grained leaf analysis of sapling trees in open-field environments. Natural scenes introduce challenges including scale variation, illumination changes, and irregular leaf morphology. To address these issues, we collected UAV RGB imagery of field-grown saplings and constructed the Poplar-leaf dataset, containing 1,202 branches and 19,876 pixel-level annotated leaf instances. To our knowledge, this is the first instance segmentation dataset specifically designed for forestry leaves in open-field conditions. We propose LeafInst, a novel segmentation framework tailored for irregular and multi-scale leaf structures. The model integrates an Asymptotic Feature Pyramid Network (AFPN) for multi-scale perception, a Dynamic Asymmetric Spatial Perception (DASP) module for irregular shape modeling, and a dual-residual Dynamic Anomalous Regression Head (DARH) with Top-down Concatenation decoder Feature Fusion (TCFU) to improve detection and segmentation performance. On Poplar-leaf, LeafInst achieves 68.4 mAP, outperforming YOLOv11 by 7.1 percent and MaskDINO by 6.5 percent. On the public PhenoBench benchmark, it reaches 52.7 box mAP, exceeding MaskDINO by 3.4 percent. Additional experiments demonstrate strong generalization and practical utility for large-scale leaf phenotyping.

Problem

• instance segmentation
• fine-grained leaf analysis
• forestry phenotyping
• UAV imagery
• open-field environment

Innovation

• LeafInst
• instance segmentation
• UAV-based phenotyping
• asymptotic feature pyramid network
• dynamic asymmetric spatial perception