Binary Classification: Determines if the leaf is from a Lemon tree.
Multi-Class Classification: If it does, this model predicts whether the leaf is healthy or affected by one or more diseases.

The models are trained using Transfer Learning from MobileNetV2, pretrained on ImageNet.

Use Cases

Early disease detection in agriculture
Educational applications in plant pathology
Research and experimentation (non-commercial)

Not suitable for real-world diagnostics without domain expert validation.

Architecture

Binary Model

Base: MobileNetV2
Input: 224x224 RGB leaf image
Output: Binary classification (Target species or not)

Multi-Class Classifier

Base: MobileNetV2
Input: 224x224 RGB leaf image (if species matched)
Output: Multi-label classification (Healthy or 1+ diseases)

Performance

Metric	Binary Model	Multi-Class Model
Accuracy	1.00	0.96
Precision	1.00	0.95
Recall	1.00	0.95
F1 Score	0.99	0.95

Datasets Used

Lemon Leaf Disease Dataset: — Licensed under CC0 Public Domain
PlantVillage Dataset: — Licensed under CC BY-NC-SA 4.0
Natural Images Dataset: — Licensed under CC BY-NC-SA 4.0

Lemon Leaf Disease Dataset was used to train both models, and images from PlantVillage Dataset and Natural Images Dataset were used as negatives for training the Binary Model.

License

This project is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license, due to the inclusion of CC BY-NC-SA 4.0-licensed datasets.

You may:

Use and share the model for non-commercial purposes
Modify it and publish derivatives under the same license
Must give proper attribution to the original data providers

Author

Muhammad Ishaque Junejo

GitHub: @IshaqJunejo
LinkedIn: Ishaque Junejo
Mail: Ishaque Junejo

Acknowledgement

Creators of MobileNetV2 Dataset Providers:

Lemon-Leaf-Disease-Dataset
PlantVillage Dataset
Natural Images

IshaqueJunejo/Lemon-Disease-Detector

Lemon Disease Detector

Overview