PlantDoc: Plant Disease Classification
> Overview
PlantDoc is a complete implementation of a plant disease classification system using a CBAM (Convolutional Block Attention Module) augmented ResNet18 architecture. Plant diseases cause significant crop losses worldwide, and early, accurate detection is crucial for sustainable agriculture. This project implements a state-of-the-art deep learning approach to identify 39 plant disease classes using dual attention mechanisms for enhanced performance.
> Technologies
> Key Features
- •Integration of CBAM attention modules into ResNet18 architecture
- •Advanced data augmentation and preprocessing with Albumentations
- •Class balancing techniques for improved performance on imbalanced data
- •Mixed precision training (FP16) and cosine annealing learning rate scheduling
- •Comprehensive model evaluation dashboards with ROC, PR curves, and Grad-CAM visualizations
> Performance Metrics
v1
accuracy:0.9746
precision:0.9921
recall:0.9917
v2
accuracy:0.9671
precision:0.9921
recall:0.9917
> Visualizations
Model Evaluation Dashboard
Dashboard showing classification examples, confidence distributions, and confusion matrix
Classification Examples
Examples of correct vs incorrect classifications
Confidence Distribution
Prediction confidence histogram for correct and incorrect predictions
Confusion Matrix
Confusion matrix across all classes
Class Distribution
Dataset class distribution pie chart
Precision-Recall Curves
Precision-Recall curves for top-performing model
ROC Curves
ROC curves for top-performing model
Training History
Training confidence and accuracy evolution over epochs
Grad-CAM Visualization
Grad-CAM heatmap highlighting disease regions
> Team
Jeremy Cleland
Graudate Student