Image recognition, voice recognition, speech recognition, and many more is just a tip of an iceberg of what a human brain can accomplish, but which is quite challenging for a computer programs, even the most sophisticated ones. As we, as humans, perfected our tools from sticks and stones to wireless networks and powerful processors the question arrives: What can be done, to teach machines to some of the basic tasks that human brain can do?
Now we are not going to dive deep in all the aspects of the topic, including Skynet and the Rise of the Machines, but we are going to take a closer look at one aspect of it. Since you have already figured out from the title of this article, it’s going to be about neural networks and their implementation in the gaming industry.
Neural Network Basics
Despite the complexion of neural networks themselves, it all goes down to a simple explanation. Human brain – the most incredible and yet not well studied computer created by nature, forms a massive network of simple calculation units that have been trained to solve complex problems, like face recognition or image recollection. This network, when simulated on the computer, is called an artificial neural network. Since engineers could recreate this biological system electronically, they figured out that artificial neural network might be trained too.
As soon as tech guys learned how to fix main challenge first computers had – not enough processing power to effectively handle lots of tasks, they have put artificial neural networks to work, using them in many aspects of our lives, including video games.
Neural networks were used in many aspects with regards to video games, creating three dimensional realities and complex worlds. The more complex video games became, the more gaming industry emerged. With the rise of streaming services and esports we have got millions of gamers all over the world. As any booming industries, gaming faces its challenges. In this article we will look at one – gamers seek better interaction, and how we can put neural networks at work to overcome it.
Play2Live Implements Interactive Tasks for Streamers Based on Neural Networks
Play2Live (P2L) – the world’s first decentralized streaming platform for gamers and esports fans, implemented algorithms for real-time monitoring of video streaming, recognition of complex objects and video content based on neural network, allowing to set interactive tasks, and monitor if the task was accomplished.
With its help, viewers will be able to set tasks for streamers choosing different conditions. Play2Live users will vote with LUC tokens – the sole mean of payment within the platform, for different tasks and set their price. For example: challenging the streamer to complete the game on the hardest difficulty level, or to use a specific weapon, equipment or skills within given period on a specific location, keep streaming for three hours straight or to start a stream on a different game, etc. This will provide new level of interaction between platform users.
How to determine whether a task was accomplished or not? The neural network monitors the stream and determines with the highest precision if the task was accomplished. In fact, the task itself is a smart contract with a deposit in LUC tokens, which is an analog of the escrow function with a deposit in LUC for the time the task is performed by the streamer.
Computer vision algorithm is used to analyze streaming images. These are neural networks, trained for recognition on their own datasets, time series and optical character recognition, including HUD (head-up display) games. Global optimization of algorithms was performed to fasten system operation time. This will allow launching the analysis of the stream even on the equipment without graphics processing unit (GPU). In future, this will also allow to perform analysis on tape drive equipment, rather than on Play2Live servers, and transmit along with the video stream meta information with results of the analysis. Implementation of such functionality will help to change the pattern of watching streaming content forever.
Interactive tasks will be developed for each popular game, and by the end of the year such functionality will be available for more than 300 games. The internal system of the neural network training will allow to add new types of events as quickly as possible. Play2Live team plans to work closely with the user community and ask fans what tasks would be the most interesting for the particular game.
Play2Live aims to combine blockchain technology with its streaming services, whilst offering 15 sources of revenue for participants – three times more compared to the streaming industry leaders. Streamers will be able to monetize their content in 11 ways versus the 4-5 available on existing platforms.