Trash management is one of Thailand's most persistent and serious environmental challenges. Thailand, being the world's tenth-largest producer of plastic-based oceanic debris, requires a system that encourages correct recycling behaviors among the Thai people.
Our project’s approach to solving this problem is to create a machine learning model that can sort between recyclables and non-recyclables; furthermore, if the object is recyclable, the model will then state which class it belongs in. Since the scope of the project has been set to objects that are commonly used and of a smaller size, home appliances and more oversized items will not be included within the dataset. The final iteration of the project will allow the user to interact with the system by inputting an image into the model through a mobile application interface. The results will help lower the barrier of entry for recycling, as the application will increase the convenience of recycling and decrease the time taken to sort the object correctly. The pictures input by mobile users can then be further analyzed to improve the model’s accuracy or developed to store and sort the classes, which can be used to tune and optimize the recycling process.
Summary and Accomplishments
We were able to construct a Flutter application to house our previously created machine-learning model. The aforementioned application accepts an image as input; the photo can either come from the user's gallery or be brand-new footage from the user's camera. Subsequently, a prediction of the picture can be made and displayed to the user, while a copy of the image is saved into an external database for future training if user consent is given. Although the current version of the app is not available in any of the stores, our prototype has been successfully installed and running on both Android and iOS devices for testing purposes.