One key feature of TensorFlow includes the possibility to compile the trained model to run efficiently on mobile phones. This enables a wide range of opportunities for researchers and developers. In this tutorial, we teach attendees two basic steps to run neural networks on a mobile phone: Firstly, we will teach how to develop neural network architectures and train them in TensorFlow. Secondly, we show the process to run the trained models on a mobile phone.
We will mainly cover:
01_Model-Training.ipynb: In this notebook, we will train an artificial neural network to classify hand-written digits based on the MNIST dataset.02_Model-Export.ipynb: This notebook loads the previously trained model and exports it as an .pb file. This enables the use of the model with TensorFlow Mobile (e.g. on Android).03_Model-Training-CNN.ipynb: This notebook shows how to train a convolutional neural network based on the MNIST dataset (i.e. same as01_Model-Training.ipynb, but with a CNN).
The compiled Android app to test the model trained in 01_Model-Training.ipynb is available in the Android Store.
This tutorial will be/was held at:
- MobileHCI'18, 3th September 2018, 09:00-17:30, Barcelona, Spain
- PerDis'18, 6th June 2018, 10:00-17:00, Munich, Germany
- MUM'17, 26th November 2017, 09:00-13:30, Stuttgart, Germany - Tutorial Website
- Mensch und Computer 2017, 10th September 2017, 11:00-15:30, Regensburg, Germany
- MobileHCI'17, 4th September 2017, 11:15-13:15, Vienna, Austria - Tutorial Website
In this tutorial, we presented simple classification examples. Beyond that, we would like to also provide plug-and-play notebooks which we used for our published papers:
- (Regression) Estimating the Finger Orientation on Capacitive Touchscreens (ISS '17): https://github.com/interactionlab/Capacitive-Finger-Orientation-Estimation
- (Regression) Estimating Finger Position on Fully Touch Sensitive Smartphones (UIST '18): https://github.com/interactionlab/InfiniTouch
- (Classification) Identifying Touches from Palms and Fingers on Capacitive Touchcsreens (CHI '18): http://github.com/interactionlab/PalmTouch