We've reached the 23rd lesson in this series! And for those who don't know, our purpose for this section is to construct a convolutional neural network in TensorFlow, which we'll train on the CIFAR10 dataset. 

CIFAR10 is a medium-sized dataset comprised of color images divided into 10 categories. What our trained model will do, when trained, is to take an image from the test set and try to categorize it appropriately. 

In this particular video, we're actually instantiating a TensorFlow session, and then train our CNN on the CIFAR10 dataset. As you can see if you're following the video, we reach a 41% performance of the model in a matter of minutes. 

This is by far a good performance, but if the model would be left to train for hours on good hardware, it could reach very good performance on this dataset, probably more than 90%. 
___
<center>https://www.youtube.com/watch?v=8oquZOFWyzI</center>
___
### <center>To stay in touch with me, follow @cristi</center>   
___
[Cristi Vlad](http://cristivlad.com) Self-Experimenter and Author