Municipal Waste Segregation Using Densenet

One of the most difficult issues facing contemporary civilization is the correct management of trash. Municipal Solid Waste (MSW) must be divided into many categories, such as bio, plastic, glass, metal, and paper. The most effective methods up to this point have been neural networks. The present deep learning methods that have been suggested to categorize garbage have been thoroughly summarized in this paper. This article suggests a system for dividing litter into the six categories listed in the benchmark techniques. Convolution neural network was the categorization architecture. These models, which Google has suggested, are based on compound scaling, have an accuracy range of 74% to 84 %, and they have been pretrained on the model. For effective categorization, this study suggests Densenet 201 model adjustment for pictures relevant to particular demographic regions. This kind of model tuning through transfer learning offers a unique classification model that is highly tuned for a specific area. Furthermore, through fine-tuning over region-specific litter photos, it led to improvised classifications.