Lakhasly

Recycling contamination is a global problem that is mainly caused by the public's confusion regarding the proper waste disposal practices in their area.Most of the current practices for educating the public on proper waste sorting include displaying labels on waste bins, distributing educational pamphlets, and accessing a municipality's website for information.Optional metadata (raw label, confidence score, category, and timestamp) is stored in a Supabase PostgreSQL database provided by Supabase for later analysis by recycling facility staff.The system utilizes TensorFlow.js and a pre-trained MobileNetV2 model that is trained on the TrashNet dataset to classify waste items on the user's device without uploading any images.This paper describes the Intelligent Waste Sorting and Recycling Assistant, a privacy-preserving web-based application that can immediately identify the type of waste an individual is trying to dispose of and provide them with the proper disposal instructions for that item in real time.They are mostly ignored by the public and do not provide the individual with immediate assistance regarding the proper disposal of a specific item at the time and location of disposal.

Recycling contamination is a global problem that is mainly caused by the public’s confusion regarding the proper waste disposal practices in their area. Most of the current practices for educating the public on proper waste sorting include displaying labels on waste bins, distributing educational pamphlets, and accessing a municipality’s website for information. However, these methods of education fail to address the problem at hand. They are mostly ignored by the public and do not provide the individual with immediate assistance regarding the proper disposal of a specific item at the time and location of disposal. Thus, there is a need for a system that can immediately identify the type of waste an individual is trying to dispose of and provide them with the proper disposal instructions for that item. This paper describes the Intelligent Waste Sorting and Recycling Assistant, a privacy-preserving web-based application that can immediately identify the type of waste an individual is trying to dispose of and provide them with the proper disposal instructions for that item in real time. The system utilizes TensorFlow.js and a pre-trained MobileNetV2 model that is trained on the TrashNet dataset to classify waste items on the user’s device without uploading any images. Optional metadata (raw label, confidence score, category, and timestamp) is stored in a Supabase PostgreSQL database provided by Supabase for later analysis by recycling facility staff. The system was tested with 25 test cases that achieved an average accuracy of 92% for the model, and the system achieved an excellent SUS score of 85.3. The system was also evaluated by 20 users who found the system to be very useful and reported a high level of privacy trust, especially regarding the responsibility the system’s developers took regarding the users’ data with an average score of 4.75/5.