Identifying Mango Leaf Diseases with Advanced Deep Learning Approaches and Convolutional Neural Networks - MangoMed AI
Tanishta, Vishesh Goyal, Dr. Megha P. Arakeri
Overview
India produces 50% of the world's mangoes. Yet diseases like Anthracnose and Bacterial Canker routinely destroy 10–39% of yields because early detection requires expert eyes that most farmers don't have access to. MangoMedAI is an IEEE-published deep learning system that detects and classifies 8 mango leaf diseases with 98.97% accuracy and an F1 score of 99.10%, using a fine-tuned EfficientNet-B0 model trained on 12,046 leaf images. Built with FastAI and PyTorch, it outperforms multiple existing approaches in both accuracy and deployment efficiency.
About This Research
The problem isn't just agricultural, it's economic. Mango Bacterial Canker alone causes 10–100% yield loss. Powdery mildew has damaged 23% of unsprayed mango trees globally. Anthracnose has caused losses of up to 39%. Farmers in the field can't diagnose these diseases with the naked eye, and getting an expert to every plantation isn't realistic at scale.
MangoMedAI addresses this with a computer vision system that identifies whether a mango leaf is healthy or infected with one of 8 diseases from a single photograph.
Model & Training
The system was built by fine-tuning a pre-trained EfficientNet-B0 on a dataset of 12,046 mango leaf images spanning 9 classes (1 healthy, 8 disease types), using the FastAI framework on top of PyTorch. The training pipeline used aggressive image augmentation like random flipping, rotation, zoom, and contrast enhancement to build robustness against real-world variation in leaf photography.
Architecture Benchmark
Three architectures were evaluated before selecting EfficientNet-B0.
| Architecture | Accuracy | F1 Score | Time / Epoch | Decision |
|---|---|---|---|---|
| ResNet-18 | 84.00% | — | — | Rejected |
| GoogLeNet | ~99.00% | — | 463 min | Rejected |
| EfficientNet-B0 | 98.97% | 99.10% | 128 min | Selected |
GoogLeNet was rejected not because it failed, but because it couldn't scale. At 463 minutes per epoch, nearly 4× slower than EfficientNet, it is impractical for deployment on mobile or edge devices.
Final Results
Evaluated across the full 12,046-image test set.
| Metric | Value |
|---|---|
| Accuracy | 98.97% |
| Precision | 99.05% |
| F1 Score | 99.10% |
| Error Rate | 0.40% |
| True Positives | 10,601 |
| False Negatives | 91 |
Why False Negatives Matter
The confusion matrix recorded only 91 false negatives out of 12,046 images. For an agricultural detection tool, a missed disease (false negative) is the dangerous failure mode, it means a farmer treats a diseased crop as healthy. This system nearly eliminates that risk.
Comparison to Prior Work
MangoMedAI outperformed all comparable methods reviewed in literature, including AlexNet-based approaches (91.2% accuracy) and ResNet variants.
Accurate, fast, and deployable — EfficientNet-B0 proves that the right architecture tradeoff isn't about chasing the highest number. It's about building something that can actually reach the farmer.
Published at the 2025 IEEE International Conference on Next Generation Communication & Information Processing (INCIP-2025).
DOI: 10.1109/INCIP64058.2025.11020297