The effective management of solid waste is still one of the most important problems, especially in urban areas where mis-separation causes damage to the environment, increased operating costs and risk factors for public health. Informal recycling workers many of whom rely on the sorting of waste for their livelihoods often work in unsafe conditions and face bleak economic prospects. Crafted in this context, recent developments in artificial intelligence and image processing bring tangible elements to make waste categorization more efficient. Here, we trained and evaluated a tailored convolutional neural network (CNN) for the automatic classification of eight recyclable waste types using grayscale images. The model used the data set containing 12,833 images were categorized into plastic, metal and glass and cardboard. Based on the SGDM optimizer methods with using of regularization techniques, it gives a valid accuracy of 96.43%, meanwhile perfectly classified almost all samples in training set. The results suggest that the proposed approach is effective in automating waste classification. More than its performance from a technical perspective, it has promising possible applications in practical environments where efficient recycling methods or working conditions & sustainable Organic Waste Management Systems within cities can be enabled.
Patiño-Ortiz et al. (Fri,) studied this question.