An innovative start-up specializing in mobile and augmented reality (AR) apps development tasked Auriga to develop an image recognition system to be used as a smart context search tool for image data for marketing and e-commerce purposes.
- 3-tier distributed architecture: mobile app, back end to handle image recognition, SQL Server Helper scripts.
- Image classification, matching and segmentation.
- Instance distribution management system and load balancer implementation.
- Aerospike as a NoSQL data storage (~1mln images).
- Work protocol implementation on top of TCP, UDP.
- Unit, functional and integration testing.
- Full cycle of code refactoring of the initial app and deployment in Amazon AWS.
- Precise real-time image classification, matching and segmentation using SVM, Chamfer Matching, and GrabCut algorithms.
- Database significant performance optimization, including cluster mode and in-memory mode configuration and adaptation.