Recyclable Waste Image Classification using Convolutional Neural Networks

Recycling waste is the way of engendering new materials and objects from the waste materials. Recycling helps us to prevent the waste of fresh raw materials and also saves energy and reduces air and water pollution. Every day, we produce waste that can be recyclable e.g. aluminium cans, glass bottles, paper, wood, and plastics etc. Dhaka City collects around 646 tonnes of plastic wastes every day, which is only 10% of all wastes created in Bangladesh. Only 37.2% of Dhaka’s plastic garbage gets recycled. Recycling waste is the only way to save this earth. In this paper, we have applied deep convolutional neural networks (CNN) for recyclable waste image classification. CNN is a type of deep learning algorithm particularly used for image classification tasks. We have created the waste image dataset using local wastes and applied four most popular pre-trained models MobileNetV2, VGG16, ResNet50 and EfficientNet for waste image classification. The experimental data show that MobileNetV2 obtained the highest accuracy 97.55% on the local waste image dataset.