-
-
Meme submission #1
-
Meme submission #2
Inspiration:
We wanted to do something that enables us to really take the experiences that we have to the next level. We understand the needs of the consumer - better tech, more realistic features and an experience thats out of this world (and takes you into another) and we didn't plan on disappointing. Here at VRSA we've taken the beautiful game and given it a tech makeover. Now instead of just watching the game, you can truly experience it... read below to find out how it all happens.
What it does:
You take a deep breadth in, the crowd roaring around you as you step up towards the ball. The referee blows the whistle and you start final play of the world cup final with all to play for.. Our simulated vision environment allows us to capture all the big games and gives the user better than front row seats to the action. However, its more than that, as it can also be the training tool that every manager has been chasing after. Our algorithms allow the match to be recreated in a virtual unity environment and the user is able to view the action from the player's perspective. Whether its the final goal of the world cup final or a friendly against Stoke, it can be used to really see the magic from the eyes of the artist.
How we built it:
To bring you the best, the team has been spent the last 24 hrs toiling away at their keyboards, exploring simulations, pushing ourselves to cover ideas we never thought possible to accomplish the brain child that is VRSA. We parallelised the work and used a deep neural network employing a combination of an object locator and an object recogniser. Unlike other neural networks, it only forwards the image once through the network, providing the speed that users always chase whilst also maintaining reliability and accuracy.
On detection, we then employed the use of GOTURN, a deep learning based object tracking algorithm which saved time on constantly having to retrieve the image components and through a combination of both we were able to accurately map our players for the duration of the match. These coordinates were then fed into a unity engine which was modelled on the environment and gave the players within it the life to move around and mimic their real world selves.
Challenges we ran into:
On starting the project, none of us had any experience with any machine/deep learning models and hence it provided a steep learning curve on taking on these challenges (one we loved). However, we had limited experience with Unity which hence provided a massive challenge. However, with our combined experience in watching Youtube tutorials online, over many hours and after many incomprehensible videos, we finally managed to understand just what was going on.
Log in or sign up for Devpost to join the conversation.