Jyothis Indirabhai, CEO

Artificial intelligence(AI) is undoubtedly the most disruptive technology of the century. Vision being the most important sensor, automating human eye is one of the key requirements for several high-potential applications. But, high compute requirements of Deep learning algorithms on pixels(video) limit the applications to mainly cloud and server based implementations. This poses a few challenges like high latency, privacy violations, connectivity issues, cabling issues and high operational cost. The edge platforms available today are very expensive and power hungry. Trivandrum based Westghats Technologies is addressing this with edge platforms that run deep-neural algorithms efficiently. The company’s strategy is to use hardware acceleration chip architecture, optimised software kernels and heterogeneous compute sharing to get the best performance out of its AI edge platforms. This brings low power, high performance and low-cost advantages for the solutions.

Westghats is a smart vision solutions company using AI as a tool to solve key problems in security & surveillance, retail and automotive market space. Westghats began its journey in 2014. The original idea was to

create a facial recognition basedcamera solution for security gates. “Initially, when we started our R&D, we came across a few hurdles. The machine learning algorithms were not matured and there was no high-performance system-on-chip (SoC) to meet the high-compute requirements of pixel processing. There were a few chips claiming hardware accelerations, but using them was very tedious and many hardware acceleration blocks in the SoCs were untested or unusable for a stable solution,” informs Jyothis Indirabhai, CEO.

"What makes Westghats stand out in the crowd is its solutions with key features, enhanced usability, minimal/no cabling requirements and high accuracy achieved through deep learning"

The facial recognition solution the team created, did not meet their expectations although this failure gave them a great idea to cash in on a new business opportunity - to create hardware acceleration architecture for AI-based video analytics. And, to build powerful and power efficient solutions powered by chips with this architecture. “We did a survey of Deep learning (emerging at that time) and computer vision methodologies and open source frameworks. We came up with a plan to build our business around this strategy,” he says.

Over the years, Westghats has been making progress in product development, solutions building and chip architecture. Westghats has developed prototypes of two smart vision platforms and the company is in the process of productising them. “While the first platform is capable for open space facial recognition and it addresses

security & surveillance and retail customers, the second platform addresses driver monitoring system for enhancing road safety for commercial fleets. Our chip architecture is complete; patents filed and are prototyped in FPGA,” he mentions.

The key SoC silicon IP cores - Neural processing unit (NPU), vision processing unit (VPU) and Image signal processor (ISP)- give good performance benchmark numbers in FPGA and are being tested for various popular neural networks used by Deep learning developer communities. “We are engaged with a few customers in Japan, US and the middle east to deploy our retail and automotive solutions,” he adds.

What makes Westghats stand out in the crowd is its solutions that comes key features to enhance the usability - minimal/no cabling requirements, high accuracy achieved with deep learning, capability to protect privacy depending on the application and low cost. “Our USP is the chip strategy to drive cost, sizeand power further down for our solutions. This will make our solutions very affordable and easily deployable,” asserts Jyothis.

Towards a Bright Future
Generating revenue through SoC IP core sales and IP enabled services, team Westghats have been supporting semiconductor and automotive Tier1 companies in the US and Japan to deliver quality products. “In Q2'2019, we will be releasing our surveillance and retail solutions. In Q4, we expect to release our fleet-safety solution. The customer demonstrations and pilot activities are in progress,” he concludes.