An Explainable AI based Plant Disease Identification using a Two-Stage Detection-Classification Pipeline
Official implementation of the paper presented at the 2025 28th International Conference on Computer and Information Technology (ICCIT).
Abstract: Plant disease looms over global food security as a significant threat. Despite this, accurately identifying diseases from images taken in real-world field conditions remains a major challenge. Standard classification models often fail in scenarios with complex backgrounds, variable lighting, and image noise. To address this, this study proposes a robust two-stage detection-classification pipeline using YOLOv11n and ECA-NFNet-L0.
Our method decouples localization from classification to handle complex backgrounds effectively:
- Stage 1 (Localization): A YOLOv11n object detector locates and crops leaf areas, achieving a mAP@0.5 of 92.9%.
- Stage 2 (Classification): Cropped leaves are processed by an ECA-NFNet-L0 framework with efficient channel attention.
Results on PlantDoc Dataset:
- Accuracy: 78.5%
- Weighted F1-score: 78.4%
The code is hosted on Kaggle for easy reproducibility. You do not need to set up a local environment if you run the notebook directly.
Click the button below to access the full training and inference pipeline:
We utilized the PlantDoc dataset for training and evaluation.
- Original Dataset: Link to PlantDoc
- Preprocessing: Details on augmentation and resizing are available in the notebook.
Coming Soon. The full BibTeX citation and IEEE Xplore link will be added here once the conference proceedings are available online.
Presented at ICCIT 2025, Cox’s Bazar, Bangladesh (December 19-21, 2025).
For questions or collaboration, please reach out to:
- Tahir Hasan: tahir.hasan.thk@gmail.com