In this bachelor’s thesis I explore how convolutional neural networks for image classification tasks can be deployed to mobile devices for evaluation purposes. I focused on models defined and trained with the popular TensorFlow-Slim library. The result of this project is an Android app that allows you to deploy such models without making changes to the source code. Moreover, I describe the differences between convolutional neural net- works and regular neural networks with regards to a core task in the field of computer vision, image classification. I then briefly outline the potential of transfer learning in the context of image classification. Finally, I give a detailed description on the design goals, development process and architecture of the app.     «

In this bachelor’s thesis I explore how convolutional neural networks for image classification tasks can be deployed to mobile devices for evaluation purposes. I focused on models defined and trained with the popular TensorFlow-Slim library. The result of this project is an Android app that allows you to deploy such models without making changes to the source code. Moreover, I describe the differences between convolutional neural net- works and regular neural networks with regards to a core...     »