Plant Leaf Disease Classification using Vision Transformer
A Vision Transformer (ViT) built from scratch using PyTorch for classifying plant leaf diseases.
The model takes a 64×64 RGB image as input and predicts one of 39 plant disease classes.
Model Performance
The model achieved approximately:
- 99.16% Test Accuracy
- Input resolution:
64 × 64 - Number of classes:
39 - Architecture: Vision Transformer
- Framework: PyTorch
Test accuracy may vary slightly depending on the training configuration and dataset split.
Model Architecture
This implementation uses a custom Vision Transformer rather than a pretrained ViT.
Patch Embedding
Images are divided into 8 × 8 patches.
For a 64 × 64 image:
64 / 8 = 8