Trash and recycled material identification using convolutional neural networks

The objective of this study is to enhance municipal garbage collection by utilizing deep learning technology and image processing algorithms to identify rubbish in public areas.  This study will contribute to the development of smart cities and better waste management methods.  Two Convolutional Neural Networks (CNN) were created to separate recyclables from landfill garbage objects and to look for trash things in a picture. Both CNNs were built using the Alex Net network architecture.  To demonstrate the approach, the two-stage CNN system was initially trained and evaluated on the benchmark Trash Net indoor picture dataset, achieving excellent results. The authors' outdoor photos obtained in the anticipated usage scenario were then used to train and test the system. The first CNN identified trash and non-trash objects on a picture database of various rubbish items with a preliminary accuracy of 93.6% using the outdoor image dataset. After that, a second CNN was trained to differentiate between recyclables and garbage that would end up in a landfill, with an accuracy of 92% overall and ranging from 89.7% to 93.4%.