I followed a tutorial on how to use TensorFlow to create digit recognition. It worked. Theres just a few issues ...

I found the code on a website, and wrote what i saw (almost copy paste)

I barely understood any of the code

I did not understand the results

I have no idea where the images was found

Im almost more confused now, than when i started, thats not good xD

https://tensorflow.org/get_started/...
You should check this out. It starts off with the most basic of tutorials.
Every block of code is commented for further reference.

Check out 3Blue1Browns YouTube channel. He explains everything so well.. here is a link. https://youtu.be/aircAruvnKk

Good luck!

@badwolf @btastic thanks. Ill make sure to check it out :)

Dude i feel you.
I have to learn tensorflow for my graduation project.
Problem is (and I cant blame them) the tutorials are designed to show you how to use their framework to implement the machine learning algorithms that you should already know pretty well.
They also seem to assume some already a priori familiarity with data analysis as they don't explain how the data is retrieved and processed etc.

So yeah
you gota learn the theory of machine learning and NN first
but I still couldn't find a nice tutorial that closes the gap between theory and practice.
So just pick your topic (let's say OCR so you probably want to use CNN) and just search for every step by step theoretical and practical tutorial of implementing that specific problem with Tensorflow.

I think, from my naive perspective, that the practical world of machine learning, as wide spread as it is becoming is still somewhat esoteric, kinda like embedded systems.

end rant

@brasorexia thanks for the reply.
I believe thats what ive tried, but i probably found the "wrong" tutorials, or just can't wrap my mind around it xD

My advice would be to start with Keras which uses TF but provides a higher level API so you can get started then move to TF when you're ready.

@DuckyMcDuckFace once my mind is ready to look into ML again, your suggestion sounds like a good place to begin with