With our solution, we have optimized neural networks on a single-board computer. This allows us to achieve server-like performance on a compact smart camera.
Video analytics in existing smart cameras cannot offer server-level performance. These cameras typically are limited to simple video analytics that are similar across different camera models. While neural networks can offer solutions for complex business issues, they require significant computer resources.
Built-in AI
SOLUTION
PROBLEM
SMART CAMERAS
A video analytics system designed from the ground up
*upon customer needs
Processing video streams on the camera via machine vision and neural networks.
SENSORS & OPTICS
*
Industrial housing for harsh conditions with 2 types of lens placement
EDGE DEVICE
Video analytics for existing video surveillance systems
ACS and automatic cost calculation, parking space control and other smart parking functions
SMART PARKING
SOON
GOALS:
Reduce equipment downtime costs.
Automate control of production operators.
Optimize production processes on conveyor lines.
Ensure occupational safety at chemical industry enterprises.
WHAT WE CONTROL:
PPE
operator vigilance
overheating of conveyor belt shafts
distributed personnel access
volume of ammonium nitrate spills
shift composition, number of employees
compliance with process regulations
CASE: INDUSTRY 4.0
RESULTS
-14% equipment downtime
-72% violations of operators
-75% falling asleep at work
-92% violation of the "red zones"
GOAL: Count incoming and outgoing passengers, their number in the bus cabin in real time using standard video surveillance cameras.
IMPLEMENTED:
Counting passengers in various visibility conditions
Integration with on-board computer for comparison with the number of validations of travel documents and calculation of the potential number of fare dodgers to increase the efficiency of the controller teams.
CASE: PUBLIC TRANSPORT
ACHIEVEMENTS:
When counting passengers using standard cameras placed in the bus, an accuracy of over 98% was achieved.
GOALS:
Automate monitoring of student and teacher attendance.
Organize collection of data on student engagement in lectures.
Ensure comprehensive security of university buildings.
IMPLEMENTED SCENARIOS:
Detection and recognition of people
Tracking head rotation and gaze direction
Recognition of open carrying of weapons
Abandoned items
Lying person
CASE: EDUCATION
ACHIEVEMENTS:
A comprehensive automatic tool for assessing and monitoring student attendance and involvement has been developed in the form of an administrative web interface and a mobile application for students; a system for preventing dangerous situations on the territory of the university's academic buildings has been organized.
GOALS:
Ensure the safety of processes in the warehouses of a beverage manufacturer.
Respond promptly to violations and emergency situations.
IMPLEMENTED SCENARIOS:
Measuring the speed of forklifts and monitoring dangerous maneuvering
Monitoring violations of "restricted zones" with dynamic changes in zone boundaries (for loading and unloading operations).
Monitoring the wearing of PPE.
CASE: WAREHOUSE LOGISTICS
ACHIEVEMENTS:
-87% total number of violations by employees. -97% PPE wearing rules violations *one month after the implementation of the CAMAI video analytics platform
Project Stages
typically 4-6 months for enterprises
Investigation
We define key success indicators, identify pain points, and offer an effective solution to the problem.
1
Neural Network Training
We record materials from the project environment, enrich datasets, customize neural network models, and further train neural networks to achieve target accuracy in a laboratory conditions.
2
Production
We select the component base, port the software, assemble cameras for the task with the necessary sensors and network interfaces.
3
Commissioning
We install the equipment on the Customer’s premises, conduct test operation, if necessary, further train the neural networks, conduct acceptance tests and put the system into commercial operation.
4
Support
We support the project at all stages. Promptly respond to requests and monitor the stable operation of the system.
5
CAMAI technology
Accelerate data processing on a single board computer, with speeds up to 30 to 50 times faster compared to running unoptimized server code.
Acquiring video stream
Applying computer vision methods
Neural network optimization
Operations minimization
Parallel processing
Established in 2004, we are an IT company that provides innovative AI-powered video analytics software solutions for improved security and safety for governments and organizations.
AB Technology
Tech Explorer Accelerator 2021 winner.
Transport Innovations of Moscow accelerator finalist.
Skolkovo innovation center resident.
highly qualified employees, including experienced developers, programmers, data analysts, and industrial designers.
Trusted by
Trusted by
Video analytics at the edge for your needs
Reduce costs associated with hardware and communication circuits
Unique solution for geographically separated video analytics systems