Video Analytics

Bell.One™ Intelligent Video Analytics Service allows to track large amounts of video data and offers the most effective and accurate response to potential threats. Video analytics involves the computerised processing and automatic analysis of video streams – whether live or recorded – to derive useful information about the content.

Bell.One™ offers an intelligent facial recognition solution that compares images of individuals from specific databases and sends alerts when a positive match occurs.

Bell.One™ solution uses a comprehensive approach to help companies stay proactive when it comes to fraud detection.

Video Analytics Security Service allows to track large amounts of video data and offers the most effective and accurate response to potential threats.

Foreign Object Identification is a set of techniques applied for the detection of foreign or undesired objects that can cause a potential hazard to the safety and integrity of society.

Use Cases

SAFE MEDIA

A comprehensive content filtering platform with minimal human supervision classifies input stream fast and accurately. It does not depend on listing or text surrounding the content or metadata. It analyzes content itself. Depending on business needs, it can do it in real time or asynchronously offering different hardware budgets.
  • Real time media channel filtering
  • Reducing human error
  • Safe visual environment for audiences

How it works

Hybrid algorithms for deep learning and image fingerprinting provide a computationally efficient method for video/image classification. Deep learning subsystem was pre-trained on 500+ hours of adult content video selected with the goal of creating representative sample.

VISUAL SEARCH

Integrate an image recognition service that identifies a replacement part in an image provided by the customer and refers him to the appropriate product in the retailer’s online store.
  • Increased sales rate
  • Reduced customer service load
  • Increased customer satisfaction

How it works

A deep convolutional network is trained on the images provided by the retailer. API for the retailer’s mobile app is provided. The customer takes photos of unknown parts. A part is identified either directly inside a smartphone or on a server depending on the size of the catalogue. The customer is directed to the appropriate item in the catalogue.