Traffic Safety Hackathon – Mark Hanson, Sony Semiconductor

Hey, everyone. Welcome back here to Tech Drug tv. I'm happy to be joined again by my friend Mark Hanssen from Sony.

Mark is the Vice President for Technology and Business Innovation. Actually, it's Sony Semiconductor, and we should get that out of the way right away. It is part of Sony, the people who bring you great TVs and audio and video equipment, but this is the part of Sony that's is specific to semiconductors.

And Mark, you could probably do a better job of explaining that than I can. Sure, no, great, great being here, Alan. Uh, yeah, Sony Semiconductors, we're, we're in probably everybody's, uh, smartphone.

Uh, we build image sensors. 5 billion of them last year. Uh, that, that's sort of our claim to fame On the semiconductor side, most recently we've established, uh, the Atrios platform and the I M X 500, which is an interesting, uh, sensor that is being utilized for, uh, ai, actually vision ai and in a way that kind of democratized it.

We talked a little bit about it previously, but, uh, because of, because of the Atrius platform and, um, our association with the Tiny ML Association, which is, uh, it's a group, a nonprofit group that works together with engineers, developers, universities, and folks to optimize AI for embedded devices. And we are one of the, uh, sponsors of Tiny ml. And this year we're actually sponsoring a hackathon, um, which is the reason I'm out here talking to you.

I want to tell you a little bit more about it, maybe get some, some of your audience to participate in this hackathon, maybe earn a little bit of money. Uh, we have some prize money associated with it, but the whole idea here is actually to work on pedestrian safety, um, throughout the United States. Say that again?

I said it's a worthy cause. Oh, yeah, yeah. Sorry, I, I didn't quite hear you.

Uh, yeah. Uh, throughout the United States, there's been a significant increase in pedestrian, uh, accidents and injuries, uh, most likely do to people using their cell phones, unfortunately, in the car when they shouldn't be. Uh, but there's these, uh, initiatives throughout the United States called Vision Zero.

Uh, the federal government supports some of these activities, and what, what the Tiny ML Association is doing is working with our local, um, uh, municipalities, happens to be San Jose, which is my hometown. Uh, and we're working with the mayor's office here to actively promote, uh, pedestrian safety by using, uh, tiny ML embedded devices like the I M X 500 and the Atrios platform to, uh, come up with solutions to alert drivers, uh, to pedestrians being in walkways they maybe shouldn't be in, uh, not just crosswalks, but mid midstream. So we're, we're working with the, the mayor's office.

They're actually part of the hackathon. They're actually doing the judging. Uh, they're, um, uh, traffic and safety light.

Uh, folks in the organization, in the mayor's office are working with us to, uh, work with some, uh, applicants who are identifying ways to maybe identify people in the lanes using our I M X 500 enabled camera, and potentially doing things like triggering, uh, overhead streetlights to flash to indicate that there may be a pedestrian or an individual in the middle of a, a roadway you might not expect. Um, but right now we're taking applications, so any of your viewers that know AI developers or app developers, uh, they can get access to, uh, hardware or, or Sony I M x, uh, hardware, uh, cameras, and can come up with potential solutions or models they could submit in and get an application and potentially win, uh, some prize money up to $3,000 today. Um, but we're still accepting applications and the program and the hackathon goes through October of this year, happens to be pedestrian safety month.

Uh, so that's when we're gonna announce the, uh, the winners. Excellent. You know, it's funny, we're, we're actually putting on a virtual event around AI later this year, I think December.

And, and we were, we had a meeting last Friday. So the, what, what do we call this event? We just don't wanna call it the AI event, right?

It's, you know, and, and one of the thing, one of the ideas concepts is, you know, separating fact from fiction, right? Because there's a, I mean, to say there's a lot of hype around AI today would be ave big understatement, right? But, you know, you have everything from this side of the, of the spectrum, which are people saying, could be the death of civilization, it's gonna be the end of all jobs, you know, all of this stuff to other people saying, you know, it's just a ni nice way of mansplaining facts pulled from the internet.

And they're not always correct that either, you know, and, and, and then as usual, the truth's probably somewhere in between, but it's good to see real world applications being explored and developed here. Right. And this is a great example of it.

You're right. We, we do have far too many traffic casualties, whether death, serious injuries in this country. And you're right, you know, pedestrian, I would say pedestrian closely followed by motorcycles, unfortunately, right.

And bicycles. Yeah. And I ride my bike down here, I scares the hell outta me.

Um, but if there's something we could do with AI to help that man, what a great use of the technology. What a great use of that technology. Yeah.

And, and we're really hoping that, you know, some clever capabilities might come out of, uh, participants. And, and this has opened up globally, so it's not just limited to the United States, but, uh, we have the ability to have applicants come from different countries. And actually, I think the last one, uh, that, that the tiny ml, uh, uh, performed the hackathon that they did, I think one of the winners was in Europe.

So, um, so, and they have a common problem as well. So, uh, the whole idea of some of these clever guys coming up with interesting models that can identify and trigger some type of an event to, to prevent these things, those are the ideal scenarios that we're looking for. And we'll, we'll send you guys the links and everything, so anybody who has interest, you Know what, we'll try to bury them in here.

Hey, mark. How, how did you get involved with the San San, um, not San Diego, excuse me, San Jose Mayor's office on this? Did they come out to you or, You know what, we, I happen to have some folks.

I work on a board of directors and we happen to have some folks who know the San Jose Mayor's office, so we kind of brought the potential hackathon to them and asked them what they thought, and they were very excited about it. They're always looking for new, innovative ways to try to implement some types of solutions. So it's an ideal way for them.

We actually introduced them to the Tiny ML Association. Uh, they have ongoing dialogue with them now, which is great. And the folks that are working on traffic signaling and pedestrian streetlights, uh, are actively involved in understanding how those organizations, universities, individual developers and others can support them in activities.

So we're really, really excited to see some of the output that might come out of that that could be helpful to the city as well as other cities, not just San Jose, but we think maybe the output could be utilized at other cities too. So, Very cool that, that, and I'm sure you will, because I mean that, look, there was a time where I sold, uh, security company. I founded a big market for us was local and municipalities and stuff.

And it very much is, I don't wanna use the word copycat, but once, once one locality is onto something, it, it kind of really spreads, right? Uh, you know, to, to others who are saying, Hey, that's a great solution to a problem. So you would hope what comes out of this does find its way, you know, not just in California.

Yeah, I think there's a lot of, uh, municipalities that get together to discuss Vision Zero initiatives. So this is just one way that we can use to get the word out and, um, hopefully, uh, enable a lot of the other municipalities to take advantage of what comes out of this hackathon. So, Yeah, I mean, good.

I'm sorry, go ahead. No, I was gonna say, I, I, I don't know how familiar you're with tiny e m l, but, uh, the organization really does concentrate on these kind of super geeky, uh, AI engineers and universities that, you know, they work with teeny little, uh, code, uh, ML code and develop these amazing models that are super high accuracy, uh, can run on these devices, many of them battery operated things like, uh, doorbell, uh, cameras and other types of implementations. Yeah.

So a lot of those technologies were initiated out of the tiny ml, uh, organization, uh, members who participated in those activities. So it's a great organization, so I love it. It really is.

And, and it, it's doing great thing. And, you know, kinda hooking AI up to these, because one of the things we're hearing with ai, right? If you're a stock market person, of course the, the companies that are creating, uh, CPUs or GNU or whatever, uh, GPUs for ai, I mean, their stocks are crazy right now, right?

But when we can marry AI to smaller footprints, if you will, I don't want to say less powerful because, you know, today's doorbell probably has a exponentially more computing power than the Apollo 10 or Apollo 11 that went to the moon, right? Yeah. I mean, Lots of things come in small packages today, but nevertheless, to be able to, to harness AI using small form factor, small footprints like that really kind of, I mean, again, huge potential there to, to really make kind of life changing, civilization changing kind of, uh, applications.

Yeah. We're seeing lots of different vendors like ourselves, uh, coming up with solutions that can, uh, help with safety, help with, uh, health environment. Um, because these devices are relatively low power, low cost, uh, they simplify the implementation of ai.

Um, and in many cases they help reduce things like, uh, power consumption. Um, you know, you got lots of cameras and edge servers running, you're, you're using lots of, uh, lots of power, which as you know, has environmental impacts. So by reducing some of those things, that enables a reduction in the impact of the environment, which is quite nice.

So a lot of these embedded devices can be put anywhere, um, run on battery. Uh, they're inconspicuous, so they can, they can be, uh, they don't take a lot of space. Um, these are relatively small devices.

So, um, those are benefits that, uh, can enable lots of different scenarios where tiny ml and even these types of hackathons can, uh, take advantage of. Sure. You know, thinking on the pedestrian safety issue, look, I, you know, my car, uh, it, it, I'm not gonna give the make and model.

I don't want to get into that, but quite frankly, could be a pain in the butt. 'cause my lane departure warning goes off all, every time I cross over. So it's obviously recognizing, you know, the white lines of your lanes and so forth.

I got, you know, and it has the hor it has so many, between all the different technologies today that our cars have from lidar for the self-driving stuff to autonomous stuff, to, you know, just some of the normal stuff that we all have in terms of sensors, I, I gotta assume we could put in there some pattern recognition for human shapes That can help us. Yeah. Most of these guys have the ability to identify humans, bicycles, scooters, uh, motorcycles, uh, as well as the actual vehicles themselves that are in the lanes.

Um, it's just a matter of these guys doing a little bit of, uh, what I call, uh, black magic under the hood to enable these really tiny models to be highly accurate in multiple types of environments, lighting, weather, et cetera. So we're Yeah, that's true too. They're very clever, very clever teams.

Yeah. So we're looking for them to kind of implement some potential solutions that these guys can utilize, but, uh, it's a common, uh, uh, uh, a common capability using tiny ml for these embedded devices. So we're hoping get lots of applicants to support these guys with their, uh, vision Zero initiatives.

Absolutely. And we'll, we'll try to embed as many links as we can get from you folks into the, uh, notes on this so people can follow along. Hey, mark, I appreciate, I appreciate what you guys are doing.

There's, there's a really cool application of, you know, of an overhyped technology, frankly, and it's good to see some real world uses, you know, being developed out of it so that it sounds great. Um, come back on, on and keep us posted or what's going on, you know, with Sony Semiconductor and the whole ATRIO system, it sounds like, you know, we're just starting to get to the good stuff. Yep.

We'll have lots of applications and partners and customers in the next few months, and we'll, we'll come back and talk to you a little bit more about it. All right. Mark Hanson, VP Technology and Business Innovation over at Sony Semiconductor on techstrong TV today.

We're gonna take a break and we'll be back here in just a moment.