An integrable system for face recognition in images and video
- Face detection and tracking in images and video
- Face comparison and search in control lists with a match score
- Face identification at different angles and tilts
- Identification of masked or partially covered faces
- Confident detection with 20 px between pupils and identification with 30 px
- Face spoofing detection (liveness) and photo suitability check for identification (QAA)
Built for large-scale projects
- 50 million face vector comparisons per secondSearching through a database of 50 million faces using 1 CPU core takes 1 second
- ½ second for face vector creation and database searchImage search takes ½ second for single requests and supports linear scalability up to X requests per second with horizontal server scaling
- Unlimited number of faces in the databaseCreation and storage of a database of people (over 100,000,000) with face photos, full names, activity history, and additional free-form text fields
- Adding cameras without expanding the system coreFor video analytics tasks, there is no need to buy a central server with extra power for future scaling, as adding cameras only requires adding video analytics servers
- Unlimited number of cameras or video filesSupport for various camera types (IP / USB) and simultaneous multi-threaded processing of camera groups
- External database for storing large imagesDelivered in Docker containers and deployed using Kubernetes in a local environment with the possibility of using an external database
Ready to integrate seamlessly
- GraphQL API and Admin web interfaceMicroservice architecture with GraphQL API via webhook and websocket
Web interface streamlines learning and configuration of key parameters - Existing integrations with video surveillance and access control systemsPartnership integrations with 4 video surveillance systems and 2 access control systems
Versatile and adaptable solution
- Support for Ubuntu and CentOSDeployment support on Ubuntu 20 and CentOS 7. Video processing client can be deployed on Windows and Linux ARM
- On-prem installation in an isolated environmentDelivered in Docker containers and deployed using Kubernetes in a local environment
- Deployment in Amazon EKS and Google KE cloud environmentsDeployment in Amazon EKS and Google KE environments with automatic scaling of hardware resources
- Edge processing — recognition without sending video to the centerEdge video stream processing prevents network congestion from high-resolution video streams
- Neural network suite optimized for low-power devicesOption to choose between full and optimized identification methods for mobile and low-power devices
- Protection against photo and data leakage from the databaseWith edge processing, face recognition is performed without sending video to the center, using only face templates without additional information, eliminating the risk of photo leakage
We accompany you through all 4 stages
Market analysis
Share our experience in similar business cases and highlight potentially risky areas.
Offer proven partnership solutions.
Solution selection
Assist in selecting the most suitable technological solutions and propose optimal architecture and equipment configuration.
Provide services for server capacity calculation and assessment of existing cameras' suitability.
Provide testing methodologies for effective participation in pilots.
Implementation and testing
Develop new functionality according to your requirements and ensure quick technical support through various communication channels, including messengers and email.
Conduct an assessment of the quality of face images in the database to predict the probability of identification errors and help configure system parameters to minimize errors.
Market entry
Implement the system under your brand (White Label) to simplify its promotion in the international market.
Offer a flexible licensing model or business model: profit-sharing, transaction-based, or based on the number of faces or cameras.
- CPU: 4+ cores 3GHz AVX support
- RAM: 24 GB
- HDD: 100 GB
Technical support
Testimonials
Save your development team's time and budget
This also eliminates the risk of inefficient organization of multithreaded processing in your project.
Client engages employees to study the technology for 2-3 weeks.
architect, developers, analyst, QA, technical writer, manager.
- 3-5 months — developing image processing functionality.
- 4-6 months — developing video stream processing functionality.
- 3-4 months — process debugging, test development, bug fixing.
- 2-3 months — optimization for different platforms and hardware.
Fast response from SDK providers is required.
Client purchases the devpack, hardware (own or in AWS/GKE cloud), cameras, and engages their developers for product integration for 2-3 months.
Client purchases the devpack, hardware (own or in AWS/GKE cloud), cameras, and engages their developers for product integration for 3-5 months.
- USB licenses are available for deployment in isolated environments without Internet access.
- Licenses can be time-limited or perpetual.
- Omni Face Recognition Platform can be licensed based on one or multiple metrics. An unlimited license is available as an option.
Idea and market analysis
Selection and fine-tuning of the neural network
For over 12 years, 3DiVi AI team has developed a robust suite of neural networks, fine-tuned to boost your project without the need to maintain your own AI team.
SDK for neural network implementation and app development
Business app developers can't use neural networks as-is. They require adaptation for different languages (C++, Java, Python, Swift, etc.) and operating systems. 3DiVi Face SDK has already done this for you.
API for accelerating integration
Work with SDK typically requires specialized developers. However, with the provided API, generalist developers can handle integration, saving money on specialized developers and reducing development time. Leverage 3DiVi Face Image API and 3DiVi Omni Stream Agent.
Templates for business logic and scenarios
3DiVi's suite of neural networks for faces, bodies, and objects allows you to integrate ready-made business scenarios (such as facial assessments, face occlusion control, identifying people crossing the lines, etc.). By using 3DiVi Face Image API and 3DiVi Omni Stream Agent, your team can avoid spending time on these implementations.
Database for events accumulating and transferring to business apps
Creating your own face recognition system can be challenging, particularly with asynchronous request processing and synchronizing face lists with the SDK's search index. 3DiVi Omni Face Recognition Platform handles this for you, leaving only the UI, roles and logs to develop.
Business logic, interfaces, roles, integrations
Your team can concentrate on high-level tasks like implementing business logic, roles, interfaces and marketing. The foundational technology is covered by 3DiVi Omni Face Recognition Platform.
Market entry and monetization