I've been thinking about a bunch of business ideas, but I've got a full time job and a community to take care of. So I'm going to share the ideas for free on this blog!
The idea is
Video analytics for competitive gaming. With the rise of competitive gaming, and the Machine Learning analytics applied to traditional sports like football, it's only a matter of time until Machine Learning analytics is applied to competitive gaming.
Once you have analytics about a game, you can use a video recording to extract highlights of the game. Who wouldn't like to have an automated highlight video based on their gaming performance?
These analytics could also be used by competitive teams to analyse opponents, reflect on past games and improve for the future games. And I believe this is where it could have the highest value and be a good business .
Machine Learning is like the chicken and the egg problem: How to build an algorithm if you don't have the data?
For that purpose, I think that it's best to start with manual prototypes as a Proof Of Concept, and then use the data to train actual Machine Learning algorithms.
For this prototype, I would advise you to focus on one game. Start by downloading a Twitch or YouTube video, and then find the areas of interest. In the following screenshot, I've chosen a video from Gotaga, where he's playing Valorant. I've highlighted some areas of interest (location on the map, score and time left, and armor and life).
You can then use a tool like ffmpeg to extract images at specific timestamps. For example you can extract an image every 5 seconds using
ffmpeg -i input.mp4 -vf fps=1/5 img%03d.png.
Then comes the Optical Character Recognition (OCR). You should manually define the zones with text to extract and run a tool like Tesseract to get the text of each frame.
Note: don't hesitate to preprocess the images to make it easier for the OCR to do its work.
From that point on, you're on your own. Optimize, improve the reliability of your algorithm, and create the best tool on the market!
Before you go and prototype, you should talk to the people that you think will pay for your tool. This part is key to make any business work and it's easy to overlook it. Especially for technical people .
Let me know what you think about the idea! Is there already a company doing this? Have you tried something similar?
This idea is free and you don't even need to mention me! I'd love to hear if you work on it though: a quick message would feel me with joy .