Facial Recognition App to Enable Retail Service Personalization
About Our Client
The Client is a big Australian company that specializes on the variety of retail services for the retail Industry. The Client created innovative services for various retail product groups (general merchandise, apparel, household appliances, grocery, consumer electronics, etc.). The Client provides individual services to adapt its solutions to the needs of each client; its solutions are trading globally and are used by many customers all over the world from single store owners to huge multi-national retailers.
Challenge
The Client decided to implement a solution for in-store service personalization and gathering customer metrics. The goal of the project was to recognize faces of the people entering stores and add information about them into the database for further analysis. As the same people enter the store again, their faces would be recognized based on the recorded data to enhance their customer experience and personalize service.
Solution
ScienceSoft divided the retail application development process into two main stages to enable:
- Face image capture
- Face recognition
Our specialists used the OpenCV library as a foundation platform to enable face image capturing and further improved its image capturing capabilities.
The second stage consisted of 3 parts:
- Preprocessing with specifically processed input images to make customers' faces recognizable.
- Calculating landmarks, i.e. some particular facial features.
- Face recognition using calculated landmarks and special criteria.
During the work on this project the team investigated a set of algorithms and methods such as Harris corner detector, Flusser affine invariants, Hu invariants, Mahalanobis metric, SIFT (Scale-invariant feature transform), etc. This combination of algorithms allowed to implement the Client 's requirements and build the system that works with a high percentage of correct results.
Results
On time and on budget, the Client received a solution that allowed to recognize store visitors using an innovative image recognition technology and tailor their in-store experience accordingly.
Technologies and Tools
.Net 2.0/3.5, C++, IPP, OpenCV library.
