Automatic waste classification using deep learning and computer vision techniques
Published by
Brac University
Summary
Authored a thesis focused on advancing waste management through custom CNN models. Trained models on a large dataset for eight waste classes, achieving high accuracy. Incorporated pretrained models (VGGNet16, ResNet50, MobileNetV2, InceptionV3, EfficientNetB0) for enhanced accuracy. Proposed YOLOv4 and YOLOv4-tiny with Darknet-53 for waste detection, achieving mAP values of 85.73% and 81.28% respectively.
