How This Startup Can Turn A CCTV System Into Video Analytics Solution

During the early 1990s, while studying for post-graduation in California, Abhijit Shanbhag observed that despite much video content being generated globally, there were almost no good video search and analytics tools available. Subsequently, as the CCTV networks all over the urban and commercial landscape started generating humongous video content -- Abhijit decided to come up with a solution that could convert the video content into machine-understandable data.

Analytics India Magazine caught up with Abhijit to understand how the startup is making deep inroads into the video analytics domain. Founded in 2016, Singapore-based Graymatics, with its offices in India and California, offers an advanced CCTV based video analytics system that can be seamlessly deployed on the cloud as well as on-premises as per the client's needs. In addition, the platform allows clients to generate critical insights from their existing infrastructure through detailed dashboards. 

"In India, Graymatics is rapidly scaling its services footprint across a wide spectrum of cases. The company offers video analytics solutions to several smart cities in India, and also in the transformation of various smart spaces such as large organisational campuses, commercial buildings such as shopping malls, highways, ports, airports and corporate clients such as Siemens, Hindustan Oil and Gas and telecom carriers," said Abhijit Shanbhag, President and CEO, Graymatics.

Was that so easy to start?

Graymatics set out on its journey with initial funding from Citrix Systems and the Singapore Prime Minister's Office. However, certain challenges were waiting for the company. Some of these included:

• The spread of cheap, low-performance and unreliable solutions in the market. This kind of below-par technology creates negative connotations about the overall market. 
• The other being peculiar to the innovation sector. Since the technology is advanced and novel for many industry verticals and users, there is some reluctance in adopting the video analytics platform beyond the early adopters. However, there is incredible and almost limitless potential as far as the usage is concerned.

Despite challenges, Graymatics recently received a strategic investment by smart city enabler Quantela. The strategic investment and alliance aim to expand Graymatics' video analytics capabilities into AI smart solutions for the target markets. "In the next five years, we plan to expand our reach across key global markets such as America, Europe and the Middle East alongside strengthening and expanding our presence in the APAC countries. We are also looking at a much larger funding round to meet our expansion plans for the years ahead," said Abhijit.

Graymatics’ USP

As a cognitive media processing company, in addition to the versatility of deployment options including edge, on-premises, cloud or hybrid edge-to-cloud, Graymatics offers:

• A cloud platform that enables the automatic analysis and recognition of products within videos and still images.
• The core products are customised AI solutions for various verticals such as digital marketing, security and surveillance, e-commerce, telecommunications, Internet-of-Things, financial sector, crowd management, smart cities, etc. 
• A highly scalable, versatile and cognitive platform that can analyse such huge video content volumes in almost real-time and turn the content into machine analysis data for future usage.
• The company offers optimised middleware to support the maximum number of high-performance AI analytics processes within a much smaller computer hardware footprint. That makes Graymatics the brand with the lowest TCO for high-performance AI systems, as the company claims.
• Other USPs include components such as a sophisticated alert report engine, forensic search, and insightful dashboards. The platform can also be seamlessly integrated with a mobile app. 

Tech behind 

The salient features of the technology behind Graymatics include:

• The company developed an in-house deep learning framework to obtain various contexts from the video/ video frames and detect objects and detailed attributes and behaviours. 
• Leverage context heavily to optimise performance. It has developed various anchor-free and NMS-free detection and advanced label assignment strategies for our prototype-to-production pipeline. It uses both Pytorch and Tensorflow in the underlying framework. 
• It leverages, co-develops with processor partners, and custom optimises various libraries for fast inference time, such as Tensor RT for GPUs and OpenVino for Intel processors.
• The architecture is developed to be highly modular, containerised with powerful orchestration tools for robustness, and with CI/CD framework for continuous upgrades.
• Use of AngularJS predominantly in front-end, Node.js in back-end and Python for application software. The company also utilises various tools for business insights, including Grafana that they have significantly customised. 
• An important element in the technology stack is a federated learning platform for CCTV for significant per-CCTV optimisation.

Wrapping up

Graymatics uses advanced AI, machine learning, IoT, and data analytics to develop futuristic video content-aware tools. The technology can perform real-time indexing, analysis and classification of videos and images as per defined parameters. Using such advanced technology, the startup can turn a normal CCTV system into a state-of-the-art and compelling video analytics solution that can generate information and insights and raise alerts wherever needed.