Artificial Intelligence Models Based on Computer Vision for Solid Waste Classification: a Proposal

This paper presents a systematic review in regards to existing artificial intelligence (AI) models for solid waste classification using computer vision, emphasizing their strengths and limitations concerning productivity parameters as well as the specifications of their training process. The aforementioned analysis aims to find potential improvement possibilities for the design of a new model and to determine the best training practices in order to obtain higher performance accuracy. This review outputs the fact that implementing transfer learning method and a hybrid dataset made-up by existing datasets, locally obtained images and image-collection methods, as well as randomly generating variations of the images by using pre-processing method can allow a model to achieve great performance stats.