Waste Classification with Transfer Learning

A computer-vision project (IBM Deep Learning with Keras & TensorFlow capstone) that classifies waste images into Recyclable and Organic categories using transfer learning with VGG16. Demonstrates an end-to-end deep-learning workflow — image preprocessing with ImageDataGenerator, feature-extraction training, fine-tuning, and evaluation — simulating a real-world smart waste-sorting system.

Proof points

• VGG16 Transfer Learning
• Recyclable vs Organic

Technologies

Python, TensorFlow, Keras, VGG16, Transfer Learning, Computer Vision

Links

• Source repository

Problem

Smart waste-sorting systems need to automatically distinguish recyclable from organic waste from images — a real-world computer-vision classification task.

Approach

Applied transfer learning with a pre-trained VGG16 backbone: preprocessed images with ImageDataGenerator, trained a feature-extraction head, then fine-tuned deeper layers. Evaluated performance and visualised predictions on held-out test images.

Results

Delivered an end-to-end deep-learning pipeline (feature extraction + fine-tuning) for binary waste classification as the final project for IBM's Deep Learning with Keras and TensorFlow course.