Built a custom convolutional neural network (CNN) to classify over 4,000 waste images into 9 real-world categories: Cardboard, Food Organics, Glass, Metal, Miscellaneous Trash, Paper, Plastic, Textile Trash, and Vegetation.
The project demonstrates deep understanding of Artificial Neural Networks and CNN architecture, focusing on practical environmental applications like smart waste management and automated sorting systems.
Tech Stack: Python, TensorFlow, GPU Acceleration, CNN, Image Preprocessing
Achieved 76.85% training accuracy and 71.64% validation accuracy on a real-world multi-class waste dataset.
Final model exported as a .keras file, ready for deployment or further tuning.