vit-base-beans
| Property | Value |
|---|---|
| License | Apache 2.0 |
| Base Model | google/vit-base-patch16-224-in21k |
| Test Accuracy | 94.53% |
| Training Framework | PyTorch 1.9.0 |
What is vit-base-beans?
vit-base-beans is a specialized Vision Transformer (ViT) model fine-tuned for identifying diseases in bean plants. Built upon Google's ViT-base architecture, this model has been specifically adapted to classify bean plants into different health categories, including healthy plants, those with angular leaf spots, and those affected by bean rust.
Implementation Details
The model was trained using the beans dataset with careful consideration of hyperparameters. It employs the Adam optimizer with a learning rate of 2e-05 and underwent 5 epochs of training with batch sizes of 8. The training process demonstrated consistent improvement, achieving a final validation accuracy of 97.74%.
- Linear learning rate scheduler implementation
- Patch-based image processing (16x16 patches)
- Transformer-based architecture for robust feature extraction
Core Capabilities
- High-accuracy disease classification (94.53% test accuracy)
- Three-way classification: healthy, angular leaf spot, and bean rust
- Efficient processing of 224x224 pixel images
- Production-ready performance metrics (F1 Score: 0.945)
Frequently Asked Questions
Q: What makes this model unique?
This model combines the powerful ViT architecture with specialized training for agricultural applications, achieving remarkable accuracy in bean disease classification. Its performance metrics, including precision (0.945) and recall (0.945), demonstrate robust and balanced prediction capabilities.
Q: What are the recommended use cases?
The model is specifically designed for agricultural monitoring and disease detection in bean crops. It's ideal for automated disease screening systems, agricultural research, and supporting farmers in early disease detection. The model can be integrated into mobile applications or larger agricultural monitoring systems.
