At the beginning of the project we searched for a
framework which meets our requirements.
Our main requirement was that we can train the
network on a desktop computer with GPU support and use the pre-trained model
on the mobile phone.
At the end, we used the Caffe framework berkeleyvision.org
In the video you can see a brief walk-through of our app. It supports the search for persons by image. The source of the image can be from contacts, file system or camera.