Gesture Detection

Smartphone motion gesture detection using Convolutional Neural Network and TensorFlow library.

Full case study in PDF

The challenge

We aimed to design a neural network scheme that is capable of recognizing desired gestures, like quick motions to the left and right. 

In order to do that, we had to pass through all steps of implementing motion gesture recognition on an Android application using the TensorFlow library. This included capturing and preprocessing training data, designing and training a neural network, and developing a test application and ready-to-use Android library.

Delivered value

We got the library that can be integrated into any other Android application to boost it with motion gestures.

The process

The idea appeared when our engineer had accomplished the first part of the "Self-Driving Car Engineer Nanodegree Program" Udacity course that partially covered machine learning. 

The process included several stages: data collection, designing a neural network scheme and training, implementation of a demo application and Android library ready to include in another projects.

You can find the full process description in the article and video demonstration on Lemberg's Blog.

Neural network design and training
Android app development
Digital media
Gesture Detection screenshot
Gesture Detection screenshot
Gesture Detection screenshot

How it works

Tensor_scheme_2 Created with Sketch. Collect Data Data Export and Pre-process Data Computer Android Device Design and Train Neural Network Use Neural Network for Gestures Recognition Optimized Neural Network Optimize and Export Neural Network Neural Network