Now Go Build - A Level Playing Field S3E4 | Amazon Web Services
- If you ask a lot of girls what they want to do when they grow up, I don't think any of them would mention, "I want to be an elite football player," because they don't see it. - Whatever the best teams in the world are getting, what can we do to bring it down to grassroots, youth level, to a group of people who have maybe been underrepresented before? If we can give that platform, that democratization, we could really redefine the sport. - Our planets and our civilizations are changing faster than ever before. Join me as I travel the globe talking to startup founders using cloud technologies to make our world more interesting, accessible, and livable. These are the entrepreneurs that are creating the future we will live in. This is "Now Go Build."
There are few things in the world that transcend boundaries and social constructs the way that sport does. Yet a gap exists between the elite and the rest of the world. Not one created by ability, but by access to technology.
The global sports industry has grown to over $500 billion in annual value. Professional organizations generate income and can afford technology to further their viewership and expertise. Yet the same tech and visibility is inaccessible to the majority of amateur athletes.
It begs a familiar question. How do we narrow this digital divide? Worldwide, athletes play sport at all levels, hoping to improve their skills, up their game, or just stay active and healthy. Would access to elite technology help them achieve parity? Could AWS technologies play a role in closing this opportunity gap? I visited Denmark to find out just that. It is a country that continuously ranks as one of the happiest and healthiest places in the world.
And there's no doubt their shared love around sports plays a huge role in that. The most popular sport here is football, and the heart of the country beats for its grassroots, youth and semi-pro teams, with over 300,000 people playing on more than 1,600 football clubs. It's no surprise, that in the rising startup community of Copenhagen, a company making elite sports technology accessible to grassroots teams was born. So you guys are both co-founders. There must be some history here. - It actually started with me being a really bad dad.
My kid was 10 years old. I was gonna go to his soccer match. I was stuck in traffic and he actually scored for the first time. You know, if I could have paid 1,000 bucks 2,000 bucks for that goal, I would've paid it. - Yeah, of course. - And I was like thinking,
this need to change, we need to develop something so other dads will have the possibility of doing that. When I got that idea, I went to the only person I know who can actually pull it off, sitting right here, which is Henrik. - So I've been playing football for the past 20 years. Like, in the lowest league.
And just when I met Keld, I just scored this amazing goal, like from the from the middle circle, right? These goals you never score in your life. - Zlatan goal. - And nobody believed it.
Because obviously nobody filmed it. So right now it's only a memory in my mind, but had somebody filmed it, I would have watched it every day. - So, but going from, let's say, your missed goals on both sides, yeah, to actually starting a company.
- The mission when we started was a more democratic way of doing sports. There's only 1% of games that are being recorded now, and that's the professional scene. That's where all the money is. But then 99% is not being recorded. - Video is actually the foundation for so many services in sports, coaching, for analysis. It's also the foundation for live streaming, for scouting, for betting.
So if you could kind of democratize the access to video, make it possible for everybody to use video in their sport, then suddenly we can make coaching better, live streaming better, the experience for the players better, and so on, and so on. We wanted to make it easy to film football matches. So there's actually two problems to be solved.
One is filming your son's goal on the smartphone, you lose the perspective and you get shaky recordings and we needed to solve, how do we find something that can actually come at an elevated vantage point. And then we said, hey, the past hundred years the way you've been filming sport, is that you have something that mechanically follows the action, that requires the camera operator. And said, what if we could turn it around and get inspired by VR, for instance, where you capture the whole scene, and then let AI, and deep learning, and cloud processing solve the rest. The cameras are obviously capturing the video. We upload that to the cloud, and this is how it looks when it arrives in the cloud. The full panoramic video that we are capturing, where we see the full pitch end to end.
Then when we have that, we then train neural networks that is able to understand the game. We're tracking the ball, detecting and tracking the players, and then we can then use that information to, one, where do we drive the camera, but then also extract data around passes, around ball position. This is then the product that the end user sees. You can not only see the match, you can also overlay the 2D map of where the players are. - People don't buy our cameras because they want a camera, because they want video. But we had to build a camera in order to get the video.
- To get the quality. - Exactly, yeah. - There's this great quote from Alan Kay that says, "Every software company that takes itself serious build its own hardware." Or should build its own hardware.
I assume someone told you in the beginning, "Do not build your own hardware." - Yeah, yeah. That's how we started out. - They told us so many times, but it was a really, really big thing for us to say, we need our own hardware.
It's the only way. - In order to get the best user experience, and the best video quality, we needed to control the hardware. So we acquired those skills, and then we developed the camera. - We needed to build a prototype. We wanted to use off-the-shelf cameras.
So we didn't needed to develop the hardware ourselves. First we started using smartphones then we started using action cameras. And what we actually did was that we took a foosball table, screwed GoPro cameras on the sideline here and then we could capture the whole scene. There was two problems. One is you had to kind of click the, you know, two buttons at the same time. When you're doing that in sync, 30 frames per second, it's impossible.
So not a great user experience. And the other challenge, we needed to upload that to the cloud to do the processing. So here we do 4K, 30 frames per second, 90 minutes. That's about 120 gigabytes of data for a 90-minute game. Uploading that in a normal clubhouse, there was just not the, bandwidth was not there. So we realized if we really want to solve this problem for our users.
We needed to do our own encoding, control the whole hardware ourselves, build a new codec that can transcode it into something, in the realms of 10 to 20 gigs that we can then upload and process in the cloud. - Understanding Veo's unique camera means understanding how everything inside plays a role in the user's journey. I spoke with Co-founder and CTO, Jesper, and the VP of Engineering, Thomas, for a deeper understanding of the tech that goes into these cameras. - When you look at a camera, you think this is just a camera with a lens, but actually this camera, it has two lenses, and it has two encoders, because when we take in the the 4K images, it's a big size, right? So we need to compress them. So we have the left and the right.
So that recording, that needs to be processed. We need to put them together generate what we call the panoramic view. - It's also something that nobody else had built before. It's not that you can buy a book off the shelf and say this is how you should build the camera. - All the cameras we've made has been calibrated, so we have like a profile of how distortion happens.
The panoramic format is a standardization that enables us to actually succeed with the models. We could numerically approximate where it should be in the model according to where the users actually see the corners. This is a key point for enabling the AI to work, because AI, with all its glory and benefits, has quite hard limitations.
- This is pretty impressive. I think the camera that you're building is amazing but that's not your real product, is it? - The use case for the customer is, they go out and they do the recording and then they turn the camera off and then they bring it home. If you just turn on the camera and you have a connection to the internet, our software will automatically detect that and it will start uploading. When the the upload is complete, up in the cloud, we can do kind of the processing and then we could generate all these kind of condensed things. So you can have a goal highlight, you can have penalty highlight. You can share that on Facebook or whatever.
We have a good idea about distribution of players. We can use that information to kind of draw heat maps for how does, you know, a team dominate. So that is super helpful because then the coach gives much more feedback to the players than if you should do it manually. The majority of our customers are not coaches that are doing the best team. They are not paid, they're doing it because they love the sport. They don't have much time.
So what we are trying to do is kind of helping them. - The model that you build to actually understand the pitch, where did you get the data from? - Initially we hoped for the best case. So we actually hoped that we could solve the problem through classic computer vision. All the tricks of the trade there, look for movements, compare pictures, look for differences. And you could get a fair bit of the way there but you cannot cut it.
We spent a fair bit of time believing that we could build a suitcase that we could give people, a laptop and a camera, go do your recording, you have a nice system, we could just work like that. But we couldn't make it work. And that pushed us into the whole machine learning area.
And machine learning means you need a ton of data and nobody can say how much. So that's back to the, going on a venture where you don't know how long it will be. - You want to develop it in the cloud before you move it over here. - Exactly.
But getting the data, getting access to a large amount of it, we're based on supervised learning, which means that we need to train the network for what we would like to get out of it. - So you have one model to create the follow cam. - Initially we started with the ideology that we need to find where the players, where the ball. Over time, we figured out it was not exactly the right way for controlling the camera.
We need to train the network for what we would like to get out of it. I think I've personally annotated like a million clicks on how a soccer ball looks like and actually manually going through to create the critical mass that enables us to run the machine learning. Doing manual control, getting motion curves over time of how the camera is moving and zooming, from that point the AI took off and actually crossed the point of we are better than humans.
- While the technological story behind Veo is a strong part of who they are, there's also a human aspect powering this technology. Coaches, players and parents at the grassroot levels are using this technology to watch, learn, and share, just like the elite clubs. To hear more about how Veo is used on the field, I met with their Head of Brand Marketing, Rob, who is a coach and also a parent.
Veo plays a role here as well? - Yeah, so what we've done is built up a really good relationship where we've almost been testing some of the tech, and working with it with some of the teams. Veo was made principally as a coaching tool to help youth, grassroots coaches who maybe wouldn't have access to that technology. We're seeing coaches use Veo for all sorts of kind of educational purposes, new ways of trying to identify and develop talent. Slowly but surely we kind of found that teams have just taken it on board and it's a natural part of how they coach, and how they watch, and how they share the footage as well. This is Japhet, he's the head coach of the under 14 Elite Reserve team. The team you've been watching show their skills today.
- Pretty educated level, not what I remember from my training days. - Thanks a lot. - So tell me a bit about the kind of things that you do here with technology like Veo. - For a coach, it's a bit of a life-changing tool that will help the player grow much faster. It's good to bring technology to them, because they're so used to it - Yeah, yeah.
- Looking at the positioning of the players, the spacing between them. When you can show them on the video, it has a much bigger impact than when you are out here on the field. When you actually can show them, pause and say "Okay, look at how much space there is between you guys." So when you go back on the field, they apply it much faster.
The kids actually asking for it to make sure that, "Can we also get the Veo during training and not just the games?" - We're seeing the things which the elite teams have, when there's been access, even at the grassroots, how quickly the adoption rate is. Very, very soon you're about to see a situation where everybody will have access. The democratization isn't just for the few, - For everyone. - it should be for many, yeah. What I'd love to do is just show you some of the session that we recorded yesterday. Here's a really good example of where we're almost behaving like an elite team, where we're able to analyze and then teach. In this instance, the players receive the ball, he performs a great pass, and then it kind of doesn't quite work as much as it should.
Now what we're able to do is go back, say, okay. We can now stop at key points. So here you're able to say, fantastic pass. But I can also draw on screen. If you had passed to the winger outside, if that winger had then played a lovely ball to the back post, and then that would've been a much easier goal.
This didn't exist three years ago. - No, no, no, no. - You wouldn't have been able to coach this at this level. This only would've been possible for the elite teams.
As a kind of amateur parent, if I can do it, then anybody can do it. - It's not just your technology that now suddenly makes this possible. 10 years ago you would have to store 25 gigs of a million games, you would've been out of your money pretty quickly. Cloud has the same democratization effect, whether you're the smallest company or the largest enterprise, technology's the same. Now you all have to compete on how good is your product, not on who has access to the best technologies. - 100%, I think we lean so heavily on that.
As the technology has improved and the ability to stream large amounts of data has become more possible, the audience just responds by eating it up. Coaches want data faster. Players want to be able to share it.
Scouts want to be able to see it. The fact that they can say, right, well you can buy our technology where it's relatively affordable, and then you can also stream in a really affordable and effective way. The democratization just gets more and more popular. And that's leading to the redefinition of the sport. - So this is where everybody comes and hangs out. The company is the intersection of sport and technology and we just wanna have it ingrained in our culture.
- So if you apply for a job here you have to be proficient in a certain sport? So it's interesting, we're talking here about the crossroads between sports and technology. Is there a way that actually technology is influencing how sports is changing? - We have now more than 20,000 clubs using our technology. We are capturing more than a million football matches. I mean we're like 50 times bigger than a professional TV crew when it comes to volume of production, right? - Yeah, but on the other hand, they also have 20 cameras.
- And they have a lot more viewers per game. But interesting here is that we are right now creating this massive base of ever growing data that the users are also annotating for us. And which means that over time, you know, there's a massive future in how can we actually give coaches insight that they would never have found themselves? Will the leagues change? Do we need 11 players on a youth team? Should a game be 45 minutes? Because now there's a technology and a format that means that it's not the federations that dictates how it is, but actually we are freeing up the sport. All the great highlights that you can watch in 10 years, creating this bank of memories. - That documentary you want to tell about the next Messi, we can actually tell that story in a way that you have never told before.
We're actually giving a lot of people around the world something that they didn't have. - Sports has enormous emotion attached to it. And whether you are the viewer, whether you're the parent, whether you're the player, the club, that makes it very attractive. - Exactly, I mean sport is a very important part of our culture. It's a glue in our society, connects people across different social layers.
Doesn't matter what car your dad drives or where you're from, it's about the game. And I think we have an opportunity for empowering the sport through technology. - The impact technology can have on sport is limitless. Veo is quickly becoming a vehicle for representation across sports. Whether the intent is to impress a scout, or simply share with friends, visibility is essential for a young athlete. Especially for one historically under-represented demographic: Women.
Veo Feminí is the company's long-term commitment to female sports. To understand more, Rob introduced me to Karoline, a Veo employee, Sports Scientist, and former youth footballer. How young were you when you started playing? - Third grade, so around nine years old. - So what motivated you in those days to start playing football? - The social aspect, to be honest. Never really thought about being a professional.
Never really thought in those lanes because I hadn't been exposed to women's professionals as a kid. If you ask a lot of younger boys what they want to do when they grow up, they're like, "I want to be a professional football player, soccer player." If you ask a lot of girls, I don't think any of them would mention "I want to be an elite football player," because they don't see it. - So how does Veo help you get on that path to exposure? - I think it's really interesting with Veo Feminí, that we're putting some more attention to the women's game. In a lot of our recordings, we see these awesome goals and highlights from women because we have just as much talent, we just don't see it. - There's been kind of a horrifically disproportionate amount of investment in the women's game over the years, but in the last 10 years there's been absolute leaps and bounds.
So one of the things which we're committed to do, that we can do, is give that platform to say, here's a place for you to show your skills, show your tekkers. - It's still male-dominated, but I'm seeing that we're going in the right direction of closing both the gap between grassroots and elites, but also closing the gap between women's and men's football. - It is undeniable that the future of sports is being shaped by technology. By tapping into the power and the potential of the cloud, Veo is providing new insights and perspectives, encouraging athletes to play more, play smarter and play together. Veo was born from a simple concept, never miss a goal.
The visibility that they provide closes the technology gap in sports, creating a more democratic playing field, where everyone gets a fair shot.
2022-12-04 07:56
