Case Study

Description

Expandable solution, for time registration, based on face recognition. Automatic determination of a person and the ability to generate employment reports. There is possible integration with third-party extensions and binding with running applications on PC. Valuable is the application which scans launched application on the desktop, bind them with individuals and generates reports.

Challenges

The team faced the task of implementing a face recognition with limited hardware performance.
Another challenge was to pick up an appropriate cluster to implement a face recognition project.
Recognition speed and overall performance was important in order to accurately calculate time without accumulating error.
Responsiveness and resistance to high loads.
Multiple face recognition support.
An important task was to determine the current launched applications and, accordingly, to log a time which user spends on an application.

Key features

• Registration of a new user (user management)
• User confirmation / merging recognition results
• WiFi access to dashboard
• Report generation
• Possibility to extend desktop applications

Approach

Embrox decided to split the project on separate sub-project, among them: back-end, front-end, the desktop listener.
The Team have created a platform, which works on desktop PC (with usage of regular USB camera) as well as on limited linux platforms (for instance, raspberry PI). Also, the system designed to works asynchronously, so it computes captured images in the background.

Results

The application has started to be used in alpha testing mode.
This solution showed that it is working stably and is possible for use as an automated logging system.

Technologies and Platforms

• Hardware platform: Raspberry Pi; module: camera
• Firmware: C/C++ Demon application
• Back-end: Go
• Front-end: VueJS
• Connectivity: WiFi
• Science: Machine Learning and Deep Neural Networking