An Automatic Environment Monitoring System Using a MobileNet Transfer Learning

Abstract

The universities have an important role to encourage and support the Sustainable Development Goals (SDGs). UIN Alauddin Makassar as one of the universities in Indonesia establishes a green campus program to support it. An automatic environment system was built to ensure the cleanliness of the university environment. A comfortable and healthy work environment is expected can improve the productivity and motivation of the academic civities. The system was built using a MobileNet architecture that using a transfer learning approach. It can detect the cleanliness level of the environment which consists of three classes: “Cleanâ€, “Less_Cleanâ€, and “Dirty†in real-time. The dataset used to train the model was obtained by capturing images of the environment around the university.  The best result of the model was achieved by using an Adam optimizer with applying a dropout in the last layer of the network. The total accuracy of the model is about 83%.