JJEM: Special Edition 2 (December 2024)
JJEM: Special Edition 2 (December 2024)
2024-12-08
Reverse Vending Machine and item verification module using ML.
Abhishek H S, Sunitha G P
Reverse vending machines (RVMs) use interactive platforms to prompt customers to return recyclables for rewards. To achieve this, RVMs require a material identification module to recognize various recyclable materials and ensure accurate payment. A vision-based detection framework was developed to identify three categories of recyclables (aluminum cans, PET bottles, and tetra packs) without the need for multiple sensors. Training and validation data consisting of 5898 user-collected samples were used to train a classification and detection framework. Pre-trained models like Alex Net, VGG16, and Resnet50 were used for classification, while YOLOv5 was employed for detection. The dataset was augmented with multiple angles and flipped images. With accurately modifying hyper parameter values (used to regulate machine learning model training), a highly accurate structure was identified. Results indicate that the detection model excels at verifying recyclable items in RVMs, surpassing the classification module.
(Reverse Vending Machine) RVM, (Field Programmable Gate Array) FPGA(controlling the operations of the machine), (convolutional neural networks) CNN.