How can I develop this ML/AI system that I want to use in my new mobile app?


I have an idea for a new mobile app. Here is what I want to accomplish using AI;
I want to get an image (png format), (maybe just byte data too), from my application (I'm developing with Unity3D/C#), send this data to AI application; get modified image from ML app and send it back to my app.

What AI is going to do with img?
Imagine you are the user of my app; you are going to draw a picture on your phone. Picture will be simple, like a seagull illustrated as an 'M' letter.
AI program will get your drawing ('M' in this case), check pixels to give a meaning to 'M'; then draw a more complex drawing that is themed around that M seagull.
(Like a drawing with an ocean made by Pixel Art, with rain clouds from Van Gogh and seagull is painted as a surreal bird...)

My general idea about building this AI system...
I'm not sure how to build this AI system, because I can't understand AI/ML completely. How it works on machine, how to implement it using computer, how to write an algorithm, how pre-made libraries like Tensorflow works... But I'm in a phase of my life that I need to use my time well, and want to build this app while learning.
I think, I can build an side app to use for analyzing and modifying image I get from user. Right know I can write in C, C# and learning JavaScript. I learned Python too (and a few others), but I'm not comfortable using it. (I hate Python). And didn't write a good program in any other language...
I thought, I can use JavaScript or Java || C++ but honestly I don't know how to start and which steps to take. Also after gaining some success I will want to port app to iOS too... Maybe that can wait...

Can you give me some examples, guidelines and advices? How to start doing this while deciding; what is best approach, performance and development-time wise?
And, Is my approach to the problem is a good one? Can you come up with a better solution for my idea that you can tell me to point me in another direction?

Yakup Murat Çaylak

Posted 2019-07-09T12:52:19.683

Reputation: 11


This is a very ambitious fist project, although you may find some pre-made neural networks that your could just wire up. E.g. or

– Neil Slater – 2019-07-09T14:17:38.877

1If you want to understand the AI/ML behind this, you should put the project to one side for a while and study the basics. It won't be possible to build your app and just add this type of component as you go along. – Neil Slater – 2019-07-09T14:18:35.523

Okay, "study the basics". I need to learn, practice, understand. I need to learn in my way, if I must say, I'm a little different on learning subject. What is the best way for this? Any resources you like, which steps to take, which subjects to look at? Also, academic side of AI/ML research is a little heavy for a solo person to learn. But I really want to learn this thing from ground up! Also I think I understand theoric side, I don't understand techical side, yet. @NeilSlater – Yakup Murat Çaylak – 2019-07-09T16:20:15.967

@NeilSlater Also thanks for links! I am checking them out. – Yakup Murat Çaylak – 2019-07-09T16:24:44.983

@NeilSlater I recently had some friends design some website for face recognition with no ML knowledge but just by using a Face recognition API. Apparently such high level apis actually exist, the OP is probably talking about Neural Style Transfer (which is popular IMO) so it might exist. – DuttaA – 2019-07-09T16:32:26.983

@DuttaA I think my idea contains both Image Generation and Neural Style Transfer techniques. Do you know such a API? – Yakup Murat Çaylak – 2019-07-09T17:32:57.403

No. but I think you can pick any piece of readymade code from GitHub (giving due credits of course) and just run it on back end. – DuttaA – 2019-07-09T17:59:57.413

No answers