Folks at DeepSystems have created a step by step 'gentle' tutorial on how to do optical character recognition with Keras and Supervisely in as little as 15 minutes (they claim).
They address a real world challenge with this tutorial: number plate recognition. The prerequisites, if you want to follow along with the coding are a Ubuntu based machine, a GPU, and docker. They provide the code on their Github profile.
Ok, so this tutorials is powered by Supervisely, which is a service by DeepSystems. In their own words:
"We at DeepSystems do a lot of computer vision developments like self-driving car, receipt recognition system, road defect detection and so on. We as data scientists spend a lot of time to working with training data: creating custom image annotations, merging our data with public datasets, making data augmentations and so on. Supervisely simplifies the way you work with training data and automate many routine tasks. We believe you’ll find it useful in your everyday work." [source]
So, you'll have to set up a free account first. After a few more prepping steps (Docker setup, dataset preparation) you will open a Jupyter Notebook with the code, and run all the cells. And that's it.
However, you'll have to go through the code yourself, step by step, to make something of it:
"Notebook consists of few main parts: data loading and visualisation, model training, model evaluation on test set. On average for this dataset training process takes around 30 minutes." [source]
At its core, the algorithm used for training is an LSTM net (Long-Short-Term-Memory). The overall architecture is, however, more complex. You can find all the details about this in the link below:
Optical Character Recognition powered by Keras and Supervisely
To stay in touch with me, follow 
@cristi
@cristiCristi Vlad Self-Experimenter and Author