cs.CVOct 3, 2026

Multi-Crop Leaf Disease Recognition: A Unified Benchmark and Cross-Region Study

Authors: Rosemary Nalwanga, Sebastian Bunda, Godliver Owomugisha, Luuk Spreeuwers, Estefania Talavera Martinez

Organizations: Data Management and Biometrics University of Twente Enschede, Netherlands · Department of Computer Engineering Busitema University Tororo, Uganda

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Deep learning models for crop leaf disease recognition routinely report near-perfect accuracy yet are typically trained and evaluated on a single dataset collected under controlled laboratory conditions, leaving their behavior under realistic cross-region domain shift poorly understood. We introduce MLD (Multi-crop Leaf Disease) dataset, a unified multi-region benchmark that combines six public crop-disease datasets from the USA, Asia, and Africa into a shared hierarchical taxonomy spanning 18 crops, 56 crop-disease classes (including one healthy class per crop) making 167,427 images. We define standardized single-source and pooled multi-source evaluation protocols that explicitly probe cross-region generalization. We also investigate whether exploiting the inherent crop-to-disease dependency via a hierarchical formulation (HiLeaD) that conditions disease prediction on the predicted crop improves recognition under cross-region shift. Under the HiLeaD, the model trained on PlantVillage achieves 99.07% in-domain disease F1 but collapses to 12.88% when tested on PlantDoc, exposing a severe cross-region domain gap. The model trained on the pooled MLD dataset partly recovers cross-region disease F1 from 12.88% to 39.64% on PlantDoc (HiLeaD), achieving a 26.76 percentage point improvement. The hierarchical formulation provides a consistent additional gain, ranging from 1.71 to 4.88 percentage points in disease F1 over the flat baseline under the MLD dataset indicating that progress in this area is currently limited more by data coverage and diversity than by model design.

Figures & tables

Fig. 1 : Image samples for diseased leaves from the different datasets where PV = PlantVillage [ 7 ] , PD = PlantDoc [ 8 ] and PM = PSFD-Musa [ 9 ] .
DatasetOriginSamplesCrops
PlantVillage [ 7 ]USA54,30614
CCMT [ 19 ]Ghana24,8814
MAK [ 20 ]Uganda46,1563
NMAIST [ 21 ]Tanzania34,3452
PSFD-Musa [ 9 ]India5,1701
PlantDoc [ 8 ]India2,56913
TABLE I : MLD dataset: a summary of the constituent source datasets.
Crop (classes)Plant VillagePlant DocCCMTMAKNMAISTPSFD MusaMLD
Apple(4)✓✓    ✓
Bellpepper(2)✓✓    ✓
Blueberry(1)✓✓    ✓
Cherry(2)✓✓    ✓
Corn(6)✓✓✓✓✓ ✓
Grape(4)✓✓    ✓
TABLE II : Crop coverage across datasets.
TABLE III : Region-wise cross-evaluation performance under different training setups. Results are noted using mean std{}_{\text{std}} F1-Score.
 PlantVillagePlantDocCCMTMAKNMAISTPSFD-Musa
CropCrop F1 Disease F1Crop F1 Disease F1Crop F1 Disease F1Crop F1 Disease F1Crop F1 Disease F1Crop F1 Disease F1
Apple99.76 0.05{}_{\text{0.05}} 99.73 2.42{}_{\text{2.42}}76.64 1.15{}_{\text{1.15}} 52.12 0.12{}_{\text{0.12}}––––
Banana––––99.83 0.08{}_{\text{0.08}} 49.38 0.19{}_{\text{0.19}}100.00 0.00{}_{\text{0.00}} 66.38 0.14{}_{\text{0.14}}
Bellpepper99.72 0.04{}_{\text{0.04}} 99.66 0.22{}_{\text{0.22}}60.03 3.70{}_{\text{3.70}} 67.95 5.52{}_{\text{5.52}}––––
Cherry99.90 0.05{}_{\text{0.05}} 99.92 0.11{}_{\text{0.11}}49.72 8.32{}_{\text{8.32}} 33.60 3.73{}_{\text{3.73}}––––
Blueberry98.37 0.89{}_{\text{0.89}} 100 0.00{}_{\text{0.00}}65.12 4.63{}_{\text{4.63}} 83.89 3.07{}_{\text{3.07}}––––
TABLE IV : Per-crop performance (mean std{}_{\text{std}} F1-score) for crop classification and disease recognition across datasets on MLD trained HiLeaD

Explore similar work

CardsList

  1. SAGE: Scalable Agentic Grounded Evaluation for Crop Disease Diagnosis

    May 10, 2026Muhammad Arbab Arshad, Tirtho Roy, Yanben Shen +7VLM EvaluationAgricultural Image Analysis

    Iowa State University, USA · New York University, USA

  2. Cross-Dataset Generalization of Bangladeshi Rice Leaf Disease Classifiers: Benchmark, Diagnosis, and Mitigation

    Sep 20, 2026Anindya PaulCross-Dataset GeneralizationPlant Disease Classification

    Department of Computer Science and Engineering BRAC University, Dhaka, Bangladesh

  3. AgriPath: A Systematic Exploration of Architectural Trade-offs for Crop Disease Classification

    Mar 8, 2026Hamza Mooraj, George Pantazopoulos, Alessandro SugliaPlant Disease ClassificationVision-Language Models

    School of Mathematical and Computer Sciences Heriot-Watt University · Institute for Language, Cognition and Computation, School of Informatics University of Edinburgh