Body Scanner

Body Scanner is a native mobile app for Android made in Java that scans the body in real time and guides the user through a series of movements of their body that are carefully tracked, recognised and counted by category.

The App uses a Convolutional Neural Network (CNN) to detect the body parts and their movements. The CNN is trained on a very lange dataset of images of the body in different positions. The CNN is then used to predict the movements of the body in real time.

This approach makes it possible to abstract the body from the background and to focus on the movements of the body independently on how the users decide to position their devices.

The app uses a model embedded within the mobile app itself without the usage of a 3rd party library or API.
This has multiple advantages: it both allows the app to be used offline and also takes advantage of the device's hardware.

The app is also built in Java and uses the Android SDK.
Tensorflow Lite is being used which allows the app to run on devices with a GPU and use the Convolutional Neural Network to its full potential using the GPU not only for training but also for inference.