Face Recognition

Product Specification

Version Q3-2021

Product Description

Incoresoft VEZHA works in security systems, safe and smart cities with large people's flows. It demonstrates good recognition results with a face size of 100x100 pixels. The minimum requirements are 75x75 pixels and can be used in malls, train and metro stations, outdoor areas. Incoresoft VEZHA scales perfectly with an increase of recognition channels. It offers continuous support, product updates, and improvements in recognition accuracy, and quality.


Face Recognition Features

  • Identification accuracy goes up to 98.35%
  • Unlimited quantity of the lists("white", "black", “VIP”, etc.) for individuals without any limitation
  • High-quality face recognition even with glasses, caps, mask, beard, and moderate facial rotation
  • Identification takes up to 0.1 seconds with a search base of 1 million people
  • Determining gender and age
  • Restrict access to notifications for a specific group of users


Camera Compatibility

The solution is compatible with all cameras, that return the RTSP stream.

A common choice for effective face recognition: AXIS  P1447 - LE,  AXIS  P1448 - LE.


Camera Installation Requirements

  • Quality: the highest
  • Bit rate: maximum possible
  • Profile: the highest possible
  • I-frame interval (GOV): 50
  • Stream smoothing: off
  • Exposure and brightness: such that for a given scene the face is clearly visible (if the camera is aimed at a light source, then an overexposed background is allowed)
  • Shutter speed: do not set too slow (more than 1/50), because in this case, the effect of blurring moving objects appears
  • Camera installation height requirements: recommended - 2.2 m (Minimum - 1.4 m; Maximum – 3.5 m)
  • People faces must be fully visible in the frame
  • There should be no reflective surfaces in the frame: glasses, mirrors, etc
  • A strong side light source (sunlight from a window) is not allowed, due to which one half of the face is overexposed


Requirements For a Video Stream

  • Optimal resolution for module operation: HD or FullHD or higher with an aspect ratio equals 16:9
  • Frame rate: at least 10 frames per second
  • Video stream must not be mirrored (horizontally reflected)

Connection

RTSP

Minimum frame rate

+

Video formats supported: AVI, .MKV, .MPEG4, .MOV, .WMV, .DVR, .ASF, .RT4, .DIVX, .264, .GE5, .TS, .3GP, .XBA (single and multistream), .MP4, .FLV+

h.263, h.264, h.265 codec support

+


Technical Specifications

Cross-platform solution - Linux, Windows

+

Centralized management

+

Ability to work with various options for video sources

+

Determination of gender and age

+

Ability to set the size of the face in the frame

+

Identification of a person with a database

+

Configuring video analytics and real-time alarms

+

Ability to process statistics and generate reports 

Microsoft Power BI and others

Face recognition module

+

Search (filterable) recognition data in the archive

+

Working with lists

+

Filtering people behind the camera and selected period

+

Notification settings: sending to the system, by Email / Telegram / SMS

+

Ability to receive reports by the hour, day, week, or month

+

API integration

+

Database import (CSV) and export support (EXCEL / PDF / СSV / JSON)

+

Integration with third-party resources - JSON messages via HTTP, MQ, TCP

+


Minimum Software Requirements

Operating System - Ubuntu 18.04 or Windows 7 / Windows Server 2008 R2

Databases - MySQL 5.5 and newer, PostgreSQL 9.3 and newer, Oracle Express Edition 10g and newer, SQL Server Express Edition 2008 and newer


Minimum Hardware Requirements

CPU

Intel Core i5-8250U @ 1.60GHz and with higher performance

MEMORY, RAM

8 GB RAM

HDD

20 GB / day

GPU


Any from the list below or higher with better performance:

GTX 1660 6GB or

RTX 2060 SUPER 8GB or

RTX 2070 SUPER 8GB or

RTX 2080TI 11GB or

RTX 3070 8GB or

RTX 3080 10GB or

RTX 3090 24GB

Bandwidth

5 Mb/s per 1 video analytic or higher*

RESPONSE TIME (LOCAL NETWORK)

200 ms or faster


Security Features

Restrict access to a specific group of users

+

Storing passwords of internal users of the system as a hash function

+

Encryption of information within the system with dynamic keys that change each time the module is started

+

Scalability of the solution

+

Business logic building system

+

Support for NVIDIA CUDA technology

+

Integration with various VMS - Milestone XProtect, Nx Meta, Luxriot Evo

+

Plugins Availability - Milestone XProtect, Nx Witness, Luxriot Evo

+

Protection of personal data

+

Support for different interface languages

+

Resolution

1080 px or higher

Detection time

8-15 ms

Recognition time

10-20 ms

Productivity

up to 500 fps in 1080p HD video

Face size for detection

minimum: 45x45 px, recommended: 80x80 px

Face size for recognition

minimum: 100x100 px, recommended: 120x120 px

 Turn

optimal up to 45º

Slope (horizontal or vertical)

optimal up to 30º, horizontal or vertical

Camera support

all camera models, RTSP stream

Maximum number of cameras

Unlimited number, depending on PC or Server

Adjusting the accuracy of filters

+

Protocols -RTSP, H264 Motion JPEG JPEG AVI

+

Search by file and comparison with the database

+

Filter change background

+

Creation of statistical data on the basis of the analyzed

+

Combination with analyzed data from third-party sources (ERP, CRM)

+


Alarms and Events

Alarms based on comparison with the list

+

Alarms with sending events to other resources

+

Alarms based on object classification: face

+

Alarms about face recognition

+

Notifications based on the accuracy of the match

+

Notifications with a given face size

+

Open tools for integration and development

+


Software Licenses

The license model works per channel (on-premise or perpetual).

SKU

NAME

LICENSE QUANTITY

IV-FR

Incoresoft VEZHA  Face Recognition

1


Face Image Characteristics

  • Illumination of faces in the frame should be steady and constant
  • If the camera is installed in front of a bright light source (the sun behind the entrance door, etc.), then it is necessary to adjust the exposure (or brightness) so the face in the frame is bright. In this case, an overexposed background is allowed
  • Image quality must be at least average
  • There should be no significant compression artifacts. No blurring of the faces of moving people
  • Image must be colored