Adaptive Domestic Waste Segregation using Image Classification

Urban India currently generates 62 million tons of waste per year; by 2030, this figure is expected to rise to 165 million tons. People rarely make arrangements to provide suitable dustbins for segregated garbage in both residential and commercial contexts. As a result, the goal is to automate waste sorting at the source while minimizing human interaction. Image classification technique is used to quickly and effectively classify waste in the trash can. Unlike object detection of a specific entity, where objects of that entity do have some comparable traits and attributes, garbage can be of any dimension, item, colour, or texture. This makes the task of detecting garbage objects difficult. The effectiveness of various pre-trained convolutional neural networks for garbage categorization is compared and tested in this study.