Thursday 15:00–15:30 in Main Track

TrashAsistant: A kivy App, which uses Deep Neural Networks, for helping trash segregation

Olgun AYDIN, Krystian ZieliƄski

Audience level:
Intermediate

Description

We have developed a mobile application using kivy framework. It uses Deep Neural Networks to help segregating trashes. Thanks to this app, before throwing trash to specific bin, you will know which trash bin you should throw to. Only thing you will do is just taking a photo of the trash via the app.

Abstract

Municipality in Gdansk started to be really strict about segregating trash last year. They are checking segregating performance of residential buildings and providing opportunity to pay less tax to those people who live in the buildings which have better segregation performance. Also, they are fining companies which don't follow the segregation instructions. Even though, instructions are well defined still sometimes people are confused to decide which trash bin they should use. For example, empty carton of milk seems like it should go to the bin for paper, but actually it should go to the bin for plastic and metals.

We have realized that having a mobile app which will guide people about this purpose would be very helpful. We have developed deep neural networks(DNN) using transfer learning. DNNs have been trained by using keras Python library. After having good performed model, kivy app has been developed.

In this speech, we would like to talk about transfer learning, Python keras library, kivy framework, obstacles we had and future plans about additional features to the application.

Subscribe to Receive PyData Updates

Subscribe