ViT Beans Classifier
This model is a Vision Transformer fine-tuned for bean leaf disease classification.
Model
Repository: Moussito/tpmlops
Base architecture: Vision Transformer (ViT)
Dataset
Dataset: AI-Lab-Makerere/beans
The model classifies bean leaf images into three classes:
- angular_leaf_spot
- bean_rust
- healthy
Evaluation
Accuracy: 0.9849624060150376
Evaluation loss: 0.08576356619596481
Usage
Use the model with the Hugging Face image-classification pipeline and the repository Moussito/tpmlops.
Example result
The published model was successfully tested on the Beans test dataset.
True class: angular_leaf_spot
Predicted class: angular_leaf_spot
Confidence: approximately 96 percent.
MLOps
This project demonstrates:
- Model fine-tuning
- Model evaluation
- Publication to Hugging Face Hub
- Model Card documentation
- Inference from Hugging Face Hub
- Deployment with Hugging Face Spaces
Limitations
This model was created for educational purposes and was fine-tuned specifically on the Beans dataset.