Machine Learning and Pattern Recognition project for classifying leaf images from three plant classes: Basil, Lemon, and Chinar.
The project extracts 13 features from RGB leaf images:
- RGB channel means and variances
- GLCM correlation features
- Template matching scores
The extracted features are used to train and evaluate three classifiers:
- Ridge Classifier
- Random Forest
- Multi-Layer Perceptron
Feature relationships are explored with pairplots, histograms, and PCA. Model selection is done with GridSearchCV and stratified K-fold cross-validation. Final performance is estimated using nested cross-validation.
The best performing model was the MLP classifier with a mean outer cross-validation accuracy of 0.929.
| Model | Mean accuracy |
|---|---|
| Ridge Classifier | 0.908 |
| Random Forest | 0.886 |
| MLP | 0.929 |