AI-Driven Classification of Bitter Gourd Varieties in India: CNN-PCA Synergy

This study looks at different types of bitter gourd in India using a mixed method with Convolutional Neural Network (CNN) and Principal Component Analysis (PCA). The set of images contains 15,000 high-quality photos of four important types of bitter gourds like Priya and Preethi. They were well chosen from many places around the world as they can be found at different stages of growth and have different shapes. Preprocessing meant making pictures the same size (224x224 pixels) and adjusting pixel values to improve model work better. The CNN idea, with layers that use convolutions and max-pooling plus fully connected parts, showed strong skill at taking in features. This gave an impressive total correctness of 97.7%. But sometimes, people get mixed up between Pusa Do Mausami and Pusa Vishesh because they look very alike. The Confusion Matrix study gave detailed information about differences in classifications. It helped improve future model designs. These results show that automated picture-based grouping systems in farming can be very useful. They provide real uses for growing crops, improving plant breeds, and dividing markets. This study builds a basic structure to use advanced machine-learning techniques for identifying types of crops. It greatly helps in improving farm technologies.