EcDNN - Video Analytics

Overview

Cipherdot’s Enterprise class Deep Neural Network (EcDNN) is developed using an open neural network architecture. It gives maximum flexibility to applications and frameworks in choosing the appropriate inference models on specific hardware and software platform to suit an organization’s workflow and culture. It uses Convolution Neural Network (CNN) – a deep learning method which provides remarkable results on face recognition in unconstrained environments. The EcDNN architecture is scalable across multiple verticals such as enterprise surveillance, city surveillance, banking, retail, traffic management, advanced object tracking, employee activity monitoring, and attendance management. Cipherdot’s EcDNN is capable of localizing and recognizing people and objects using a single neural network.

Cipherdot’s EcDNN addresses some of the most common problems faced by enterprise organizations today:

  • Security personnel find it difficult to keep track of multiple entities.
  • Tracking unauthorized persons and objects, especially in sensitive areas.
  • Subsequent real time detection and identification of authorized personnel.
  • Productivity analysis.

Solution Overview

  • Real time image streaming using high resolution 4K cameras.
  • Face recognition with > 97% accuracy.
  • Augmented training-based CNN with improved facial recognition rate.
  • Real-time image comparisons and alerting mechanisms.
  • Efficient tracking of unknown entities.
  • Efficient video analytics for people & object movements.
  • Unauthorized access control.
  • High performance user-friendly software with 24/7 monitoring capability.
  • High quality custom reports with Statistical Analysis.

Product Features

  • Easy to use dashboard interface.
  • Interface can be customized and integrated into existing security system.
  • Real-time analysis, tracking and counting data based on user definable areas.
  • Accurate Face detection and matching system.
  • High performance software reducing the need for unnecessary hardware.
  • Powerful onsite or remote management.
  • Custom reports for in-depth statistical analysis.
  • Automatic scheduled reporting delivered via email.

Customer Benefits

  • Easy to deploy: Customizable as per customer requirements
  • One-stop-shop: Complete in-house platform, coding, patents & algorithms
  • Save time and money: Scalable solution requiring less time and supervision
  • Highly accurate facial recognition: High detection accuracy and noise resistant
  • Analytics: Define, view, and export movement statistics
  • Continuous surveillance: Uninterrupted round-the-clock monitoring
  • Hassle free auto-labeling feature for facial recognition to shorten training phase on custom datasets.
  • Auto-labeling feature may be extended to objects with asset tags.

Competitive Advantages

Advantage

Cipherdot EcDNN

Competitors

Model Size

194 MBs

270 - 300 MBs

Execution Speed (1-5 faces/frame)

45 fps

5 - 25 fps

Inference Time

0.022 seconds

0.040 - 0.2 seconds

Accuracy

79 mAP @ 40 fps

60 - 75 mAP

 

74 mAP @ 50 fps

 

Accuracy on LFW benchmark

99.38%

80% - 99.8%

Facial detection & recognition

Simultaneous

Sequential

Network Architecture

Single with multi-tasking feature

Multiple neural nets dedicated for separate tasks


Custom Training Dataset Creation

The following shows the automated methodology of creating customized datasets of the objects where the model has to perform.

Real-time Object Detection

Cameras transmit real-time video to cloud server following RTSP which ultimately help in detecting the objects in the processed video.

Product Use Cases

Employee Productivity Management

According to a recent study by Gallup, only 13% of the employees worldwide are engaged in their jobs. Employee productivity is key to enterprise organizations because it helps organizations understand their operations and processes better. Understanding employee productivity also helps organizations improve efficiency of its workforce. Cipherdot’s Video Analytics provides non-intrusive employee management that will help organizations eliminate the need to maintain cumbersome employee work history and attendance systems.

Prevent Tailgating

Tailgating is one of the most innocent security breaches in organizations. While employees are only being courteous when holding a door open for others, this may lead to unauthorized access for visitors without badges, or the passive acceptance of an uninformed worker. There are certain sensitive zones in organizations where authorized employees are required to swipe his/her personnel tags to ensure his/her presence is recorded. However, some find it convenient to tailgate another employee who has gained access to the area by swiping his/her tag. Unauthorized entry allows individuals who could intend harm to property and employees. Cipherdot’s Video Analytics will help maintain a database of all employees in the organization. Reports of any unauthorized access will be immediately sent to administrators, who can then verify whether the person has the necessary approvals to be present in the building.

Prevent Workplace Harassment

Recent statistics show that at least 1 in 3 women, and 1 in 5 men are subject to harassment at the workplace. Almost 20% of workers say they have experienced discomfort due to sexual humor. The non-intrusive approach to employee management that EcDNN provides will help prevent unwanted advances that some people are subjected to at the workplace. The EcDNN architecture will help HR admins in organizations to implement policies that are best suited to both employees and the organization.

Unauthorized Access and Security Breach Control

Data centers and equipment storage rooms are sensitive areas in any organizations. Any security breaches into these areas may prove expensive to the organization. Instead of spending hours and days investigating when such breaches happen, the EcDNN architecture will immediately alert the responsible authorities about such incidents, thereby securing such sensitive areas, and saving crucial dollars for the company.

Other Potential Use Cases

Cipherdot’s Video Analytics solutions finds extensive benefits outside enterprise applications. The EcDNN architecture may be extended to airport surveillance, traffic management, retail management, banking, etc. The architecture may also be used in video based home security systems.  According to recent statistics, 65% of home burglaries happen between 6am and 6pm – time when people are at work. The average loss per burglary is $2,230. And 34% of the burglars enter through the front door. While home automation and alarm systems greatly help in alerting the owner and the authorities about such a crime, it may not provide direct evidence of the crime committed. Cipherdot’s Video Analytics architecture will help keep track of the residents in a gated community, or an independent home, and help provide direct proof to authorities and the residents when such a crime happens. Hence, the time required to investigate the crime is drastically reduced, which will ultimately result in reduced crime rates.