Data labeling using YOLO image anotations
Developed a dataset labeling pipeline for a smart waste management system that classifies garbage into categories like plastic, organic, metal, and paper using computer vision. Collected and curated a dataset of 8,000+ images from open-source datasets and manually captured images under different lighting and background conditions. Designed a labeling workflow using tools like LabelImg to annotate bounding boxes and assign category labels. Performed data preprocessing including resizing, normalization, and augmentation (flip, rotation, brightness variation) to improve model generalization. Trained a Convolutional Neural Network (CNN) model using TensorFlow/Keras to classify waste images. Achieved ~87% validation accuracy. Integrated the labeled dataset with a simple backend API (Node.js + Express) to allow real-time image uploads and predictions. Built a basic frontend interface to test classification results. Tech Stack: Python, TensorFlow/Keras, OpenCV, LabelImg, Node.js, Express, React