🌿 HerboNet

📝 Notes

This project was developed to address the critical need for accurate identification of medicinal plants in Ayurveda. By leveraging transfer learning with state-of-the-art CNN architectures, HerboNet provides a digital tool that can accurately classify plant species from images, supporting practitioners, researchers, and enthusiasts in the field of traditional medicine.

📄 Project Description

HerboNet is a deep learning-based image classification system designed to identify Ayurvedic medicinal plants from photographs. The system uses Convolutional Neural Networks (CNN) with transfer learning to achieve high-accuracy plant recognition across a diverse dataset of indigenous flora.

• 📊 Dataset: Initially sourced from Kaggle containing 40 plant classes with 120 images each, augmented with additional samples for 16 new plant classes to enhance diversity.
• 🧠 Model Architecture: Multiple transfer learning models evaluated — EfficientNet, MobileNet, and ResNet — with EfficientNet achieving the highest accuracy and selected for final deployment.
• 🔧 Preprocessing Pipeline: Image resizing, normalization, train-test split, and data augmentation to improve generalization.
• 🎯 Classification: Accurately identifies the name of an Ayurvedic plant when an image is uploaded.
• 📈 Performance Comparison: Rigorous benchmark of model performance across architectures to select the optimal deployment candidate.