Fill form to unlock content
Error - something went wrong!
Your content is just a step away. Please submit below.
While a few cities in the world are beginning to share their streets with driverless cars, the rest of us are stuck in traffic with all the other humans behind the wheel. And these drivers may or may not pay attention to the rules of the road or look out for pedestrians and bicycles.
Fortunately, smart cities are taking important steps to improve these matters (while we wait to be bumped into the backseat), and AI and video-management solutions play a big role in that transition. The data that cameras collect provides a crucial window into driver and non-driver behavior, and the analysis of that data can lead to powerful insights and solving real-world problems—like snarled traffic and tragic accidents.
But there are many applications for AI video analytics in contexts beyond traffic jams—from personal protective equipment detection to retail—as Srivikraman Murahari, Vice President of Products and Strategic Alliances at Videonetics, a video management, video analytics, and traffic management provider, explains (Video 1). He also addresses using partnerships to create end-to-end solutions, the privacy and security concerns around collecting all this data, and the power of AI video analytics to affect our everyday lives.
What are the challenges that AI video analytics can help address in urban planning?
One of the major challenges that government officials and city planners see is in citizen noncompliance with traffic rules, and this can cause accidents and even fatalities. So that puts pressure on government officials to streamline the traffic situation and smooth the traffic flow.
There are now 100-plus smart cities where our Videonetics intelligent traffic-management solution is deployed, and it has very powerful analytics capabilities with traffic. We also provide a smart visualization for the government officials, which gives a lot of insights for taking further action. I can confidently say that the traffic flow in those 100-plus smart cities is smoother and more streamlined now, and more awareness is created among the citizens to adhere to the traffic rules.
What are the challenges to implementing AI video analytics in smart cities?
One challenge is field of view—with cameras, the field of view is restricted. We are exploring methods such as sending drones to difficult places to capture the video. So we are looking at many innovative ways for the cameras to reach difficult areas.
How do you implement this technology while balancing citizen privacy?
That’s a good question. I would say that we have to enforce responsible and collaborative AI. And when I say collaborative AI, I mean that the government officials, the independent software vendors like us, and the citizens should all know what is happening, should know how the data is getting used. There should be a very transparent data policy. The second thing I would say is to use minimized, anonymized data. That means not storing so much data, and then the data that is stored should be anonymized.
At Videonetics we have very, very strict security standards. For us everything is objects, and we don’t have any people data. We abide by international security compliances, and we have very strict standards in our protocol and the way we handle the data. We are transparent and ensure data safety and compliance with those international standards. That’s how we handle it, and I think these are my suggestions.
“Adopt #data and #technology—including responsible and collaborative technology, and responsible and collaborative #AI” – Srivikraman Murahari, @videonetics via @insightdottech
Can you provide some examples of deploying AI video analytics in a smart city?
As I mentioned, we’ve deployed our platform in 100-plus smart cities, and it has helped to smooth and streamline the traffic and to ensure the safety of citizens. For smart city we are the number one in India. I can talk about a case study in one of the premier cities in India.
In that city there are about 400 cameras monitoring the traffic and another 700 cameras in pipeline—so I’m talking about 1,100 cameras monitoring the city’s traffic, ensuring lane discipline and one-way traveling, etc. It has eased the operations of the administrators in smoothing traffic flow.
As far as implementation is concerned, we have a collaboration with all the leading camera vendors around the world. For each project we decide on the most suitable camera along with the systems integrator and the partner who is involved in that project. The analytics then happen on the edge. For edge we use the Intel platform extensively—the Intel® Core™ i5, i7, i9 series as well as the latest generation chipsets, 11-13. And then in certain scenarios we have cloud for the storage.
Coming to the question of how to do it efficiently, our R&D is continuously putting in effort on that; we have a dedicated effort into how to optimize the compute. And I can say we have traveled a long way on that from the time we started. Now we have, let’s say, 20x or 30x improved computing efficiency. We are looking at how to use many fewer frames of the video to deduct an event, instead of processing the entire video. We are looking at collaboration with partners and using their latest technologies, platforms, and solutions to optimize performance and computing powers.
What are the benefits of partnering with other companies like Intel?
The partnership with Intel has been very great, very exciting because we are focusing more on and traveling more in the direction of analytics on the edge. And that’s a direction that Intel is also promoting—more analytics on the edge, more analytics by CPU. And so Intel is our best, our top partner in this direction, a direction that matches both the organizations.
Secondly, we have used Intel’s OpenVINO™ platform—the OpenVINO deep-planning platform. That enhances models using techniques such as post-training optimization and neural-network compression. These things reduce the total TCO for the customer because the computing power is enhanced. Another very great thing to mention about Intel is the Intel® DevCloud platform, which is always available for us to benchmark our latest models. As we speak, our models are getting benchmarked in the 11th to 13th generation series of Intel chipsets.
And I’m very happy to announce that we won the 2023 Intel Outstanding Growth ISV Partner Award for outshining the competitors, as well as enabling Intel to onboard more partners. So it has been a very long and successful journey with Intel for us.
What other use cases for AI video analytics can we look forward to?
Outside of smart cities, we are into quite a number of verticals. The biggest space there is aviation and airport security, where we are helping more than 80 airports in analytics, such as being able to quickly detect smoke and fire. And then there are enterprises such as oil and gas and thermal—and smoke and fire are pretty dangerous there, too. These kinds of video-analytics applications have been quite a hit and create a lot of value for these enterprises.
We have our own deep-learning platform, called Deeper Look, with about 100 video-analytics applications developed out of it. They cover a wide range of analytics, including crowds, vehicles, mass transport, women’s safety, and retail. In retail we do a heat map that gives owners insights to help them understand selling patterns in their stores. In the case of mass transport, most of the railway in India is using Deeper Look. Another very widely employed use case is PPE detection, which helps with the safety of workers. There is also banking and finance. One other interesting area we support is forensic research, which is very useful for investigation.
Any final thoughts or key takeaways?
My key takeaway would be to adopt data and technology—including responsible and collaborative technology, and responsible and collaborative AI—to increase the vigilance of governance, to increase the operational efficiency of enterprises, to enhance the safety of people, and to go beyond security.
Regarding the computing, we have to continuously invest in optimizing the computing power; we have to be open in our API; and we have to show a lot of openness so that our platforms are easily interoperable with third-party vendors. That is also quite important.
And finally, I repeat: Ensure responsible and collaborative AI, and take the administrators and citizens into confidence. Video and IoT are an excellent combination, and there can be lots of use cases that will enrich the quality of human lives.
To learn more AI video analytics in smart cities, listen to AI Video Analytics Empower Communities: With Videonetics. For the latest innovations from Videonetics, follow it on Twitter and LinkedIn.
This article was edited by Erin Noble, copy editor.