AI-Powered Waste Sorting Data Labeling & Model Training
I developed a real-time waste classification system that involved training an AI model to identify and categorize waste items from images. The process required collecting, annotating, and labeling a custom dataset of over 6,500 images across multiple waste categories. I applied transfer learning techniques and evaluated model performance for accuracy improvements. • Collected and annotated images of waste types such as plastic, metal, and paper. • Used YOLOv4 models and OpenCV for object detection and image classification tasks. • Labeled and verified image data to improve model reliability. • Integrated labeled datasets into an embedded system optimized for Raspberry Pi deployments.