Material Stream Identification System

Developed an end-to-end Machine Learning system designed to automate waste material identification and classification from images, supporting smarter waste sorting and recycling processes.

The system classifies waste materials into seven categories: Glass, Paper, Cardboard, Plastic, Metal, Trash, and Unknown. It leverages a pretrained EfficientNetB0 deep learning model to extract multi-scale image features, combined with traditional machine learning classifiers, including K-Nearest Neighbors (KNN) and Support Vector Machine (SVM), for accurate material classification.

Implemented a complete machine learning pipeline covering dataset preparation, data augmentation, duplicate image removal, feature extraction, model training, hyperparameter optimization, and performance evaluation. Additionally, integrated an unknown material detection mechanism to identify unfamiliar or unsupported inputs, alongside real-time webcam inference for live waste classification.

Technologies: Python, TensorFlow, EfficientNetB0, Scikit-learn, OpenCV, NumPy, PCA, KNN, SVM.