FastAI sounds like a good starter.
I’m not sure what you mean by “all deep learning codes are so small”.
Do you mean that the code does not contain many lines of code?
If so, I wouldn’t underestimate the energy which was used to get these (few) lines of code to a working model.
. I am doing the course, but as interest in ai , other friends interested in other things wrote alot of coding in project. I feel in just 30 60 lines of code, i would build a whole deep learning model in just 1 day.
I like and loving this AI journey. But as because of small code , everybody would learn AI, which can be dangerous in our career, if it is easy . Also we dont really have to understand the mathematics .
At the risk of saying out loud what some forum participants might be
thinking, your speculation that you can build a deep-learning model
in 60 lines of code and that learning AI might be easy sounds a bit
Sure some easy (toy) problems are easy, but some hard problems
are hard, and some harder problems are hard enough that nobody
knows how to solve them yet.
For a reality check, you might look at some of the public datasets,
and some of the deep-learning competitions that use them. As one
example, you could look at NIPS: