As SOFTECH celebrates its 20th anniversary, we also had the honor to launch two new artificial intelligence apps at the leading technology shows in Europe: MWC in Barcelona and CeBIT in Hannover. MoneyCam is the AI app dedicated to the visually impaired people who need help in recognizing bills when they go out shopping or paying their invoices. We thought that computer vision technology and deep learning algorithms may be of great help in solving a practical issue.
MoneyCam Features
– with MoneyCam AI app Euro money bils may be recognized on mid-range Android smart phones
– the development team defined Sound instructions for helping visually impaired with money placement and recognition
– vibration mode is alos available when the bill is in the right position
MoneyCam AI app by SOFTECH has a voice assistant that guides you to correctly position the bill
MoneyCam Technical Details
– The MoneyCam Android AI application was built with the purpose of being efficient – it works on cheaper, mid-range phones – and in offline mode (!) so that no internet connection is required in order to make it available for a larger audience
– MoneyCam Android AI application is built on top of Tensorflow Mobile platform by using Tensorflow Object Detection API
MoneyCam AI Computer Vision Technical Details
Deep neural nets training, evaluation and deployment:
– The Data-set: Using only two images per bill – front and back, MoenyCam Data-Set Generator creates enough data in order to let the neural network learn to recognize the banknotes in different viewing and lighting conditions;
– The Models: Different MobileNet v1 ConvNet models are being used for both the detection and classification of the bill. Google designed these models for efficiency on mobile platforms.
– Bill Detection: To detect the bill placement in the camera’s view a Single Shot Multibox Detector is employed that provides real-time detection using the lightest version of MobileNet v1 models as feature extractor.
Image sources: we used http://machinethink.net/blog/googles-mobile-net-architecture-on-iphone/ and SSD Multibox paper: https://arxiv.org/abs/1512.02325
MoneyCam AI app at MWC 2018
The AI app was launched at the Mobile World Congress 2018 in February. It was well received by the visitors at the stand as it was found useful and innovative. SOFTECH from Cluj-Napoca was among the few exhibitors from Eastern Europe who showcased own developed apps in the Artificial Intelligence area.
MoneyCam AI app at CeBIT 2018
In June 2018, it was time to get the pulse on our AI app once again. MoneyCam was featured at CeBIT 2018 in Hannover too! Again it was a success, since its underlying technologies may be applied to other type of uses.
Further developments
Based on MoneyCam AI application model, we can decline several types of useful applications for other domains too.
Should you have specific needs or imagine a similar solution for education programs, for research, or other related fields, feel free to write us on contact@softech.ro
