Year of Graduation
Recyclable Waste Classification Based on Machine Learning
System and Software Engineering
The research concerns waste classification problem using CNNs and smartphone camera for on-device classification. The waste is classified into 6 categories: cardboard, glass, metal, paper, plastic, trash. The dataset chosen is formed from TrashNet and openrecycle datasets. The neural network model selection is based model’s inference time and accuracy. As a result of the experiments MobileNet model was chosen, it achieved 91% accuracy on the dataset. The model was then converted to CoreML model and was implemented on iOS mobile device for real time classification.