Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 26 min ago
Peer-reviewedCC BY 4.0Frontiers in Plant Science

MSPRP: multi-granularity symptom-aware probabilistic representation prompting for target-domain-free plant disease detection

  • Xian Cao1,2
  • Sixiang Quan3,4
  • Yalu Zhang1,2
  • Jun Liu1,2
  • Xianlong Tang3,4
  • Tai Zhang3,4
  • Zhihui Wang1,2
  • Yuankun Wang5
  • Yingying Peng1,2
  • Huali Li6
  • Xiaoxia Xiao1,2
6 affiliations
  1. 1School of Informatics, Hunan University of Chinese Medicine
  2. 2AI TCM Lab Hunan, Hunan University of Chinese Medicine
  3. 3Second Surveying and Mapping Institute of Hunan Province
  4. 4Hunan Engineering Research Center of Natural Ecosystems Carbon Sink Monitoring
  5. 5Department of Geography & Geographic Information Science, University of Illinois Urbana-Champaign
  6. 6School of Artificial Intelligence, Hunan University
Open at sourceDOICite

Abstract

Introduction: Existing vision-language zero-shot anomaly detection methods have mainly been validated on industrial datasets and often suffer from lesion-scale variations, domain shifts, and complex background interference when directly applied to field plant images. To address these challenges, this paper proposes MSPRP, a multi-granularity symptom-aware probabilistic representation prompting method for target-domain-free plant disease detection.

Methods: Based on frozen CLIP encoders, MSPRP introduces three modules: the Multi-Granularity Symptom-Aware Module (MGSAM) performs learnable hierarchical fusion and spatial-channel symptom attention on multi-layer patch features, the Uncertainty-Aware Prompt Flow Module (UPFM) models healthy/diseased prompts as probabilistic representations to improve prompt adaptability, and the Adaptive Cross-Modal Prototype Aggregation Module (ACPAM) aggregates suspected symptom prototypes from high-response patches to reduce background interference in image-level disease scoring. The model is trained only on the industrial MVTec AD dataset and tested on target agricultural datasets without target-domain training.

Results: A field Polygonatum odoratum disease dataset is also constructed as an evaluation benchmark. Experimental results show that MSPRP achieves the best average AUROC and AP among methods that do not use target agricultural training images. Ablation results verify the contribution of each module.

Discussion: Overall, MSPRP provides a target-domain-free computational methodology for field plant disease discrimination without target agricultural training images or target-domain pixel-level lesion annotations.

Related

  1. Multi-granularity Adversarial Generation Integrated Consistency Representation for Chest Low-Contrast-Enhanced CT Synthesis
    Lin Zhao, Shangwen Yang, Dianlin Hu +5
    IEEE Transactions on Medical ImagingPeer-reviewedCC BY-NC 4.0Full textSystems & synthetic biologyHAL
  2. Enhanced Early Detection of Allergic Rhinitis: A Prospective Study on a Symptom‐Based Predictive Model
    Ke‐Zhang Zhu, Chao He, Si‐Zhe Zhu +2
    World Journal of Otorhinolaryngology-Head and Neck SurgeryPeer-reviewedCC BYFull textImmunologyDOAJ
  3. WSD-YOLO: A lightweight YOLO-based model with enhanced feature representation for maize pest detection
    Shenming Qu, Zhiheng Liu, Huazhen Zhao +2
    PLOS OnePeer-reviewedCC BY 4.0Plant biologyPLOS
  4. Dual-domain self-supervised feature alignment via spectral–spatial representation learning for deepfake anomaly detection
    Yi He
    PLOS OnePeer-reviewedCC BY 4.0Full textBioinformatics & computational biologyProbabilityPLOS
  5. Enhancing Clinical Classifier Robustness with Deep Representation Learning Based OOD Detection on DNA Methylation Data
    Md Zahangir Alom, Quynh T. Tran, Breuer Alexandar +1
    arXivPreprintCC BY 4.0Full textGenetics & genomicsarXivPMC

Cite

BibTeXRIS
BibTeX
@article{cao2026msprp,
  title = {{MSPRP: multi-granularity symptom-aware probabilistic representation prompting for target-domain-free plant disease detection}},
  author = {Xian Cao and Sixiang Quan and Yalu Zhang and Jun Liu and Xianlong Tang and Tai Zhang and Zhihui Wang and Yuankun Wang and Yingying Peng and Huali Li and Xiaoxia Xiao},
  journal = {Frontiers in Plant Science},
  year = {2026},
  month = oct,
  doi = {10.3389/fpls.2026.1921532},
  url = {https://www.frontiersin.org/articles/10.3389/fpls.2026.1921532/full},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - MSPRP: multi-granularity symptom-aware probabilistic representation prompting for target-domain-free plant disease detection
AU  - Xian Cao
AU  - Sixiang Quan
AU  - Yalu Zhang
AU  - Jun Liu
AU  - Xianlong Tang
AU  - Tai Zhang
AU  - Zhihui Wang
AU  - Yuankun Wang
AU  - Yingying Peng
AU  - Huali Li
AU  - Xiaoxia Xiao
PY  - 2026
DA  - 2026/10/29
JO  - Frontiers in Plant Science
DO  - 10.3389/fpls.2026.1921532
UR  - https://www.frontiersin.org/articles/10.3389/fpls.2026.1921532/full
AB  - Introduction: Existing vision-language zero-shot anomaly detection methods have mainly been validated on industrial datasets and often suffer from lesion-scale variations, domain shifts, and complex background interference when directly applied to field plant images. To address these challenges, this paper proposes MSPRP, a multi-granularity symptom-aware probabilistic representation prompting method for target-domain-free plant disease detection.  Methods: Based on frozen CLIP encoders, MSPRP introduces three modules: the Multi-Granularity Symptom-Aware Module (MGSAM) performs learnable hierarchical fusion and spatial-channel symptom attention on multi-layer patch features, the Uncertainty-Aware Prompt Flow Module (UPFM) models healthy/diseased prompts as probabilistic representations to improve prompt adaptability, and the Adaptive Cross-Modal Prototype Aggregation Module (ACPAM) aggregates suspected symptom prototypes from high-response patches to reduce background interference in image-level disease scoring. The model is trained only on the industrial MVTec AD dataset and tested on target agricultural datasets without target-domain training.  Results: A field Polygonatum odoratum disease dataset is also constructed as an evaluation benchmark. Experimental results show that MSPRP achieves the best average AUROC and AP among methods that do not use target agricultural training images. Ablation results verify the contribution of each module.  Discussion: Overall, MSPRP provides a target-domain-free computational methodology for field plant disease discrimination without target agricultural training images or target-domain pixel-level lesion annotations.
SN  - 1664-462X
ER  -