The target group appreciated the application. Soon the wish appeared to develop a version for the iPhone, as this is a widespread device among blind and visually impaired people.
Patrick Valenta, a student at the University of Applied Sciences Augsburg, developed a version for the iPhone as part of his bachelor thesis. In addition, the recognition of the pedestrian lights and the traffic light phases was completely revised and is now based on the artificial neural network "YOLO", which was trained with more than 3,000 images of pedestrian traffic lights, collected by volunteers throughout Germany. As a result, the accuracy of the recognition, as well as the speed of the evaluation, improved significant.
The advancement of the app was only possible due to the generous support of the Reinhard Frank-Stiftung, the PRO RETINA-Stiftung and the Dr. Gabriele Lederle-Stiftung.
The app is available in the Apple AppStore for free. Since its release, the app has been installed almost 400 times and is now used 30 times a day on average.
The app is now also used by rehabilitation trainers for the mobility training of blind and visually impaired people.
For the near future, it is planned to revise the Android version of the app, so that it also benefits from the use of the artificial neural network for detection. Photos of pedestrian traffic lights have been collected by volunteers during the past few months. With the help of this, the detection should be further improved and made even more reliable.
Those who would like to participate and support the project can do this easiest by photographing pedestrian traffic lights. For this purpose, a project was submitted for participation at www.buergerschaffenwissen.de, which will be launched in the next few weeks.
Furthermore, in cooperation with Nikolauspflege Stuttgart, a study will be carried out to objectively evaluate the use of the app and to collect feedback for the further improvement and development of the app.
More information about the app "Ampel-Pilot", the support of the project and the source code, which is available as open source, can be found on the homepage of the project at http://www.ampel-pilot.de.