Last year, Prisma AI, a global company providing visual AI-based solutions, partnered with Adani airports to set up ‘Desk of Goodness’ where the company looks to provide swift assistance to passengers in need by alerting their support staff through monitoring.
“The concept is to aid passengers. Anybody carrying a baby, or walking with a crutch, or if someone falls, we want to capture these situations and trigger help to our on-the-ground support staff, who would be having a tablet to see the visuals and the location of the passenger,” said Amitabh Chowdhury, executive director and COO, Prisma AI, in an exclusive interaction with AIM.
Not to Worry, Your Data is Safe
The AI for Humanitarian Goodness project at Adani airports deals with vast amounts of private data, where their camera captures information and sends it to a server for analysis. Once the server detects something significant, it notifies a ground staff member via a tablet. However, the notification is not sent through the most preferred service provider.
In other words, Prisma AI has set up an independent notification server at Adani to bypass Google’s control and ensure connectivity.
“90% of the world’s notifications on your phones go through Google servers only. Google has a monopoly on notifications. They [Adani] clearly said that they will not give internet access. So we have our own notification server, which we implemented for them over there, and those notifications are only going in through their local Wi-Fi network.”
The ‘desk of goodness’ service is available across six airports, including Lucknow, Ahmedabad, Mangalore, and Trivandrum.
Apart from airport service, Prisma has primarily serviced security, finance, infrastructure, and road assistance, including name plate recognition. The company has built its proprietary computer vision technology under the name ‘Gryphos.’
Inside Gryphos
Gryphos serves as the core computer vision platform for the company. It utilises deep convolutional neural networks, enables comprehensive analysis of videos, images, objects, faces promoting analytics and predictive capabilities. Prisma has over 100+ global deployments across five continents with over 21k cameras running Gryphos.
“Gryphos has various core engines, and feature engines, and over the years, we have started building derivative products out of these core engines, something like ‘Veri5,” said Chowdhury. The product is a face authentication system which is finding use in financial institutions and security solutions.
The accuracy of results via their platforms are set according to the sectors they cater to. “For instance, in a banking transaction, I cannot afford to make a mistake, where face recognition is required to proceed to the next step. So. I will set the cutoff at a higher percentage, say 75%. Whereas, if a police is searching for a lost child, I set it at 30%. It doesn’t matter if they get five or 10 different lost children. I don’t want to miss out on that one truly lost case. So, the use case defines the use data parameterisation,” said Chowdhury.
Evolution of Prisma AI
In the late 1990s, Prisma was originally headquartered in Germany. In 2017, the company was acquired by Prisma India, and the parent company’s headquarters moved to Singapore. However, Prisma Global India has a core development team that works out of Mumbai.
Prisma’s project, assisting Interpol in Germany to track down paintings and artefacts stolen by the Nazis towards the end of World War II, was among the first use cases.
Prisma AI has also worked on a project to piece together East Germany residents’ torn and shredded letters that were kept in 16000 gunny bags. Only when the Berlin Wall came down did this problem come to light, and the National Archives requested it be put together.
“It was this whole exercise of a jigsaw puzzle. One bag might have had a thousand-odd letters. They kind of set up a conveyor belt system and put up a camera at one end, and each of those pieces was snapped. Our algorithm kicked in to try and match and put it together. This was kind of the genesis of visual AI if I can call it in those days, and since then, we have moved on to many things,” recalled Chowdhury.
Walking the AI Talk
Unlike many other companies in the space, Prisma AI is known for successfully applying AI in real-word scenarios.
“Unfortunately, even till today, 90% of AI projects around the world do not succeed, and, it’s more of an academic interest, to a large extent, for many of those guys. So, the Googles, and the Amazons, may have made it open source, but those codes are not really being applied that much in the real world scenario,” he said, believing that these are instances where Prisma emerges as a differentiator.
Chowdhury also observes a major shift in how businesses approach AI of late. “I don’t have to sell AI, as a concept, as much as I did, say, five to eight years ago. The moment I say the word ‘AI,’ everybody is interested,” he quipped.
In January, Prisma AI announced its partnership with India’s Pro-Kabaddi League team, Jaipur Pink Panthers. The collaboration will improve the in-stadium fan experience and streamline and enhance venue security. The company is also working with a few partners in the US and Mexico to bring a similar experience to stadiums.
