Neural networks
ComBox Technology is actively developing the application of neural networks in business and the public sector. As a technology stack, we focus on training networks managed by Tensorflow with further conversion and optimization for Intel OpenVINO, which allows us to significantly reduce the operating costs of using innovative solutions.
Almost any fast-growing industry implements neural networks as a tool for automating and optimizing processes. On the basis of analytical systems, we implement complex solutions for the functioning of many processes without human participation. We organize effective management of infrastructure facilities and help ensure security in various areas of activity, both for private and state customers.
Neural networks applications
-
Detection of static and dynamic features of cars
Field of application-motor transport, automobile traffic. Recognition of vehicle registration plates, model / brand, color. -
Container or shelf full detection
Scope - shelves, refrigeration equipment or containers for the collection / storage of materials. Automatic tracking of space filling in containers or on flat surfaces (in%) and smart notification. -
Passenger and pedestrian detection
Scope – public transport. Passenger traffic is recorded using head detection and trajectory analysis for tracking held objects and calculating the direction of movement. -
Face recognition system
Scope-face recognition in a crowd or in a small group of people. Biometric identity verification and collection of other analytical information.
Implemented detectors and classifiers
- Vehicle plate detection and recognition
- Vehicle type detector
- People detector, people counter
- Face detector
- Passenger counter on transport
- Face recognition (face recognition by 5/30 points of the face)
- Classifier gender / age by faces
- Smoking detector - https://habr.com/ru/company/intel/blog/493726/
- Medical mask detector for staff. Telegram bot demo - https://tlgg.ru/imasks_bot
- Infectious disease symptom detector for staff - https://habr.com/ru/post/499874/
Some examples
