Lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillation

Real-time, on-device segmentation of rubber tree powdery mildew is crucial for smart agriculture, yet existing deep learning models often fail to balance accuracy with computational efficiency. While large models can be precise, their high complexity hinders deployment on terminal devices; conversel...

Full description

Saved in:
Bibliographic Details
Main Authors: Yuheng Li, Lisha Ye, Yu Zhang, Lijuan Zhou, Qian Chen
Format: Article
Language:English
Published: Elsevier 2025-12-01
Series:Smart Agricultural Technology
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2772375525005507
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849228333840072704
author Yuheng Li
Lisha Ye
Yu Zhang
Lijuan Zhou
Qian Chen
author_facet Yuheng Li
Lisha Ye
Yu Zhang
Lijuan Zhou
Qian Chen
author_sort Yuheng Li
collection DOAJ
description Real-time, on-device segmentation of rubber tree powdery mildew is crucial for smart agriculture, yet existing deep learning models often fail to balance accuracy with computational efficiency. While large models can be precise, their high complexity hinders deployment on terminal devices; conversely, smaller models often struggle to identify small, multi-scale disease spots in complex backgrounds. To bridge this gap, this study develops and validates light-PMSeg, a lightweight instance segmentation framework designed for accurate, real-time disease grading. Our contributions are fourfold: First, we establish a high-performance baseline, PMSeg, which integrates a Mixed Aggregation Network (MANet) and a Channel-wise Cross Fusion Transformer (CCFT) to significantly enhance sensitivity to small targets and multi-scale feature extraction. Second, to achieve on-device feasibility, we systematically compress PMSeg into light-PMSeg through a novel combination of slimming pruning and knowledge distillation (KD), which restores performance after parameter reduction. Third, we constructed a rubber tree powdery mildew segmentation dataset PM-Dataset containing 6,200 images. Finally, we propose an automatic grading algorithm based on the segmentation masks to provide actionable disease severity levels. Experimental results demonstrate the superiority of our framework. The final light-PMSeg model achieves a mAP50 (Mask) of 90.1% and an F1-score of 85.4% at a speed of 148.6 FPS. Crucially, this is achieved while reducing the model's parameters by 79.6% compared to the unpruned PMSeg, confirming its suitability for deployment on low-storage devices. This study delivers a validated, end-to-end solution that effectively reconciles high segmentation accuracy with a lightweight architecture, presenting a practical and robust method for intelligent powdery mildew monitoring.
format Article
id doaj-art-e5a189a2e723425ba274e5a6aeb302f2
institution Kabale University
issn 2772-3755
language English
publishDate 2025-12-01
publisher Elsevier
record_format Article
series Smart Agricultural Technology
spelling doaj-art-e5a189a2e723425ba274e5a6aeb302f22025-08-23T04:49:52ZengElsevierSmart Agricultural Technology2772-37552025-12-011210131910.1016/j.atech.2025.101319Lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillationYuheng Li0Lisha Ye1Yu Zhang2Lijuan Zhou3Qian Chen4School of Cyberspace Security (School of Cryptology), Hainan University, Haikou 570228, China; Key Laboratory of Internet Information Retrieval of Hainan Province, Haikou 570228, ChinaSchool of Cyberspace Security (School of Cryptology), Hainan University, Haikou 570228, China; Key Laboratory of Internet Information Retrieval of Hainan Province, Haikou 570228, ChinaSchool of Tropical Agriculture and Forestry, Hainan University, Danzhou 571737, China; Sanya Institute of Breeding and Multiplication, Hainan University, Sanya 572025, ChinaSchool of Cyberspace Security (School of Cryptology), Hainan University, Haikou 570228, China; Key Laboratory of Internet Information Retrieval of Hainan Province, Haikou 570228, China; Corresponding authors at: School of Cyberspace Security (School of Cryptology), Hainan University, Haikou 570228, China.School of Cyberspace Security (School of Cryptology), Hainan University, Haikou 570228, China; Key Laboratory of Internet Information Retrieval of Hainan Province, Haikou 570228, China; Corresponding authors at: School of Cyberspace Security (School of Cryptology), Hainan University, Haikou 570228, China.Real-time, on-device segmentation of rubber tree powdery mildew is crucial for smart agriculture, yet existing deep learning models often fail to balance accuracy with computational efficiency. While large models can be precise, their high complexity hinders deployment on terminal devices; conversely, smaller models often struggle to identify small, multi-scale disease spots in complex backgrounds. To bridge this gap, this study develops and validates light-PMSeg, a lightweight instance segmentation framework designed for accurate, real-time disease grading. Our contributions are fourfold: First, we establish a high-performance baseline, PMSeg, which integrates a Mixed Aggregation Network (MANet) and a Channel-wise Cross Fusion Transformer (CCFT) to significantly enhance sensitivity to small targets and multi-scale feature extraction. Second, to achieve on-device feasibility, we systematically compress PMSeg into light-PMSeg through a novel combination of slimming pruning and knowledge distillation (KD), which restores performance after parameter reduction. Third, we constructed a rubber tree powdery mildew segmentation dataset PM-Dataset containing 6,200 images. Finally, we propose an automatic grading algorithm based on the segmentation masks to provide actionable disease severity levels. Experimental results demonstrate the superiority of our framework. The final light-PMSeg model achieves a mAP50 (Mask) of 90.1% and an F1-score of 85.4% at a speed of 148.6 FPS. Crucially, this is achieved while reducing the model's parameters by 79.6% compared to the unpruned PMSeg, confirming its suitability for deployment on low-storage devices. This study delivers a validated, end-to-end solution that effectively reconciles high segmentation accuracy with a lightweight architecture, presenting a practical and robust method for intelligent powdery mildew monitoring.http://www.sciencedirect.com/science/article/pii/S2772375525005507Powdery mildewGrading segmentation networkModel pruningKnowledge distillation
spellingShingle Yuheng Li
Lisha Ye
Yu Zhang
Lijuan Zhou
Qian Chen
Lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillation
Smart Agricultural Technology
Powdery mildew
Grading segmentation network
Model pruning
Knowledge distillation
title Lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillation
title_full Lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillation
title_fullStr Lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillation
title_full_unstemmed Lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillation
title_short Lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillation
title_sort lightweight grading segmentation network for powdery mildew based on model pruning and knowledge distillation
topic Powdery mildew
Grading segmentation network
Model pruning
Knowledge distillation
url http://www.sciencedirect.com/science/article/pii/S2772375525005507
work_keys_str_mv AT yuhengli lightweightgradingsegmentationnetworkforpowderymildewbasedonmodelpruningandknowledgedistillation
AT lishaye lightweightgradingsegmentationnetworkforpowderymildewbasedonmodelpruningandknowledgedistillation
AT yuzhang lightweightgradingsegmentationnetworkforpowderymildewbasedonmodelpruningandknowledgedistillation
AT lijuanzhou lightweightgradingsegmentationnetworkforpowderymildewbasedonmodelpruningandknowledgedistillation
AT qianchen lightweightgradingsegmentationnetworkforpowderymildewbasedonmodelpruningandknowledgedistillation