Date of Award
Bachelor of Science
Department Chair or Program Director
Major or Concentration
Americans recycle 32.1 percent of all the waste they create, as confirmed by the latest report from the Environmental Protection Agency. However, the underlying issue that remains is that most Americans are not equipped with the knowledge of the correct methods of recycling – and the deficiency of that knowledge is the greatest quandary to ensuring that the country is “green” and environment friendly. A survey of two thousand American citizens revealed that 62 percent of them worry that this inadequate knowledge is causing them to recycle incorrectly. The aim of this research is to develop an Android App, where with the use of one’s smartphone camera, a person can capture an image of an item that they wish to recycle and the output displayed on the app will be either plastic, glass, metal or garbage(unknown). The first component of the research investigates whether a usable model can be built that can be used to robustly identify recyclable objects. A deep convolutional neural network (CNN) is built in Python and trained on a labeled data set of a thousand product images from various perspectives, to determine whether the object that is to be recycled is composed of plastic, metal or glass. In order to provide the most efficient approach, I experimented on well-known deep convolutional neural network architectures. By implementing transfer learning and fine tuning to the pre-trained models with a common data augmentation strategy ResNet101V2 model provided the best result with 82% test accuracy. A larger data set is required to reduce overfitting and increase the accuracy. The chief purpose of this project is to develop an application based on a deep learning model that aids users to correctly identify the nature of objects that they deem recyclable and to widen the scope of healthy recycling - of household and domestic goods, which is a tiny but indispensable and effective individual step that needs to be taken in order to combat pollution and in the long run, prevent climate change while there is still time to do so.
Kandel, Pratima, "Computer Vision For Recycling" (2020). Student Research Submissions. 379.