Emerging video analytics company Flixsense are looking to evolve the way sports coaching and broadcasting is undertaken through their intelligent video platform.
At the heart of it is their development of artificially intelligent video analysis that can be trained to recognise people, movement and objects that adds big insights to highly tactical sports such as football and cricket.
A pivot in late 2016 to develop their AI to analyse sports has resulted in an array of clubs and organisations wishing to learn more about the potential of the product.
The recent winner of the KPMG Game Changers Global Startup Competition at the 2017 Sports Analytics Conference in Melbourne, Flixsense co-founder Jalaluddin Shaik spoke to Bullpen about his company’s move into sport and the potential power of their technology.
Bullpen: What is your background that led you to starting Flixsense?
Jalaluddin Shaik: “I’ve always been interested in broadcast media, I love designing and building video platforms having worked on all levels of video infrastructure and delivery.
“With video there is always much more to the story than what is being shown, which demands background context and supporting media to truly enjoy it. Like when you are watching an interesting television show, and you are intrigued by the object in the background or by the actor playing the part, that you want to know more about them right away.
“Utilising my video background, I explored what was possible. That’s when I thought what if machines could watch and understand videos like we do, you could ask anything you want to know about everything in that video and that’s how the idea started.”
BP: Can you briefly describe what the technology is?
SJ: “We use deep neural networks (DNN) to build our AI pipeline to understand video. DNN allows you train highly specific aspects of interest like faces, objects and actions which you want the machine to understand.”
BP: How did you trial and iterate your idea so it understood and recognised players, logos and objects?
SJ: “The sports analytics version of the product is a recent development. Prior to this we were doing celebrity recognition and linking more context information around them like their IMDB profile, Instagram account, Twitter account. The product got some good feedback but we realised the product would lose its sheen very soon, so we thought this might not be a wide application going forward and we switched to sports in late 2016. We started training the AI to recognise players and since we did that we had a lot of interest from sporting teams to say, “if you can do that can you also show patterns, movements and tracking of players.” That’s a problem because it’s currently done manually.”
BP: Did you work with any coaches or sports scientists to assist with working on player movements or patterns?
SJ: “Once I had the first prototype I went to Football Federation Australia and showcased what I had and they told me, “we do a lot of manual coding, what we would like to do is recognise fouls in games.” That’s when the ball started rolling, I quickly realised it was not just FFA who had this problem and then we started having active interest from other teams who requested player tracking and positioning.”
“We are working with performance analysts and coaches to determine aspects which are important for them to track.”
BP: In your pitch at the Sports Analytics Conference it was very football focused, what made you consider football?
SJ: “My co-founder, James Melvin, is a very big football fan. We needed a start and football is the world game, so we thought we’ll try it but we didn’t know if it would work.”
BP: How do you think your product could really transform sports coaching and sports analysis?
SJ: “My primary objective is to build a smart AI that can watch a lot of videos and understand what’s happening. For sport this is a big problem because there’s a tactical advantage to knowing your opposition, and then consider how you would counter it, where should your best players play, which is the best line up for the day and who is in best form.
“There is only so much a human mind can analyse, elite sporting teams are very time poor and usually remember one or two things of the many recommendations analysts make after analysing tagged footage, but the machines are so powerful they could run through a whole archive of footage and understand so much more about each team, player, positions and crunch them to give concise, clear and actionable insights which matter.”
BP: Is it getting tested with other A-League or local football clubs?
SJ: “We’re currently in talks with some teams across the A-League, AFL and cricket for pilots.”
BP: What could the product do for cricket in terms of coaching and technique?
SJ: “Cricket is a very strategic game, batting, bowling and fielding are three aspects you can track and drive analytics on.
“With our product we can tell what shot a batsman plays and also track the pitch of the ball. If we link the two, which is a shot a batsman plays based on where the ball is pitched, we can help position where the fielders need to be.
“The product has lot of potential to improve grassroots cricket too, students and semi-professional players can now have access to advanced analytics just by uploading their games to our platform.”
BP: Even different pitches offer different types of batting and bowling strategy as well. From there, could you take this overseas, even going back to India and work with Indian Premier League franchises or the BCCI to test the product?
SJ: “Cricket is religion in India, with a billion followers it is probably our biggest market for cricket analytics, so certainly we have plans to reach out to the cricket ecosystem in India.”
BP: Now that you’re doing a lot of testing and trialling for a couple sports, is there any other sporting vertical that you may target?
SJ: “Not at the moment. We want to do football and cricket really well before moving elsewhere.
“We’ve had interest from other sports such as Australian Rules football, basketball and netball. Resourcing is an issue when going after all sports, so we need to pick the right ones first and look to scale quickly.”
BP: Was there a moment when someone validated your idea and you thought you have a viable company?
SJ: “At the start you have to constantly keep validating your idea. It’s an ongoing process, you’ve always got to keep meeting people and testing your ideas and know when to pivot. Having good mentors is key too, the right mentors will keep in you in check on executing the right strategy.”
BP: You have to be quite nimble as well and be willing to listen to good advice.
SJ: “Exactly. You are always listening, it’s all vital for validation. You’re listening to whether people like certain features, what issues they might have, how will they use it, how can you make it better.”
BP: How many people are working at Flixsense, what’s the size of the company?
SJ: “We have three people at the moment after hiring our first lead engineer. Now that we are part of one of Sydney’s leading tech accelerators at Startmate, and have some funding into the company, we are looking for some more like-minded people to join us.”
BP: As your product is still in the pilot phase, do you have an idea when you would be able to do a market launch?
SJ: “The current pilots are very promising. I’m hoping for a launch by mid-November.”
BP: Your AI can go across several other verticals but the one that is really interesting to me is for your AI to recognise brands – such as jerseys, LED in-stadium signage, apparel – how well is it developed to move into that vertical?
SJ: “The good thing about our AI is that it could be applied to anything in the sporting ecosystem, not just coaching, Analysis can drive broadcaster revenue through branding, advertisement and drive more engagement with the content.
“One thing we can do apart from coaching and analysis is how to do fan engagement better. Recognising players, brands and jerseys, allowing fans to receive real time insights and quick highlights from the game is good start. I think this piece of the puzzle is more interesting for broadcasters.
“Another important aspect I am exploring is talent scouting which is a huge problem. That is one of the things we will launch with our cricket analytics solution. The way that works is any player or club can upload their video, their technique can be analysed, reviewed and shared with scouts.
“It can potentially cut down a lot of leg work for talent scouts and a lot of travel going to venues solving the issue of manpower and never having enough scouts to go to different venues to scout players and cover games.”
BP: Just finally, what are your goals or challenges are you able to tackle the coming 12 to 18 months?
SJ: “We’ve had strong interest from lot of teams and governing bodies and there is a genuine problem in the industry with manual coding and analysis of games. The need for our product is potentially huge, the bottleneck for us today is to scale my team quickly to build awesome sports analytics products for all sports.
“We will need to raise funds to do that, we will open our seed round to help us scale on the back of Startmate Accelerator demo days planned on 9th and 11th October in Sydney and Melbourne.”