In this tutorial from the neural networks series we're going to look into OCR, or optical character recognition. Over the course of the next few videos we're going to work on building and OCR engine that will recognize handwritten digits. 

First things first though. We're going to use a public dataset from Stanford (available [here](http://ai.stanford.edu/~btaskar/ocr/)). We download and extract it. The Python libraries that we're using in this video are os and sys (from the standard library), numpy, and cv2 (an open source library for computer vision). 

We use a variable to store out dataset filename and we set a few parameters that we'll use with cv2 (a resize factor, the height and width of each sample, etc). Then we're gonna open our datafile and go through each row or each line (which represents the features for one digit). 

We're going to use cv2 and the 'imshow' method to visualize (inside a loop) the characters from the dataset. This visualization helps us get a better idea of what we're going to do next. We'll have to train an algorithm (a neural net in neurolab) to learn the features and the corresponding labels for each character in the dataset. 

After the learning process is complete, we asses the performance of the neural net. I guess it's much harder to simple narrate or describe what we're about to do in a programming tutorial, which is why I encourage you, if you're interested, to watch the video below for the full walk-through: 

<center><iframe width="560" height="315" src="https://www.youtube.com/embed/kpOFr7C4yJc" frameborder="0" allow="autoplay; encrypted-media" allowfullscreen></iframe></center>
___
### <center>To stay in touch with me, follow @cristi</center>   
___

[Cristi Vlad](http://cristivlad.com) Self-Experimenter and Author