Deep Learning-Based Automated E-Waste Classification Using Convolutional Neural Networks

Priya Atul Raut, Dr. Ayesha Siddiqui

Abstract: Electronic wastes have increased rapidly due to the rise in electronics. As such, there is a need to sort and recycle this kind of wastes. Manual classification and sorting of the e-wastes is inefficient and inaccurate. This has made it necessary for deep learning algorithms to be employed. In particular, CNNs will be utilized in the process of classification and sorting of the electronic wastes. In this experiment, large volumes of images representing the electronic waste have been incorporated. Normalization and resizing of the images are conducted in order to improve their quality before carrying out any further operations. Dimensions, colors, and brightness of the images have been varied to enlarge the size of the training data set. Convolutional, pooling, and fully connected layers have been employed. Besides, some other hyperparameters such as learning rates, batch sizes, and epochs have been adjusted to yield better results. It has been demonstrated that the automatic system has worked efficiently and accurately in sorting and classifying different types of electronic waste. Accuracy, precision, recall, and confusion matrices have proven the effectiveness of the classifier.

Keywords: electronic waste recycling, deep learning classification, CNN image sorting, waste image recognition, automated waste management