Convolutional Neural Network for Classifying Yoga Poses

I finished the Deep Learning Specialization on Coursera! Obviously, Andrew Ng has a track record of being a fantastic instructor, and it’s a great introduction to the entire field and includes lots of fun projects. However, you realize by the end that you still have a lot to learn! So, hours later, I embarked on my first deep learning project — building a simple convolutional neural network (CNN) with Keras for classifying yoga poses.

I wanted to build everything from scratch. I did almost immediately find a knowledge gap. All of the courses that I’ve taken come with nice, ideal data sets and do a lot of the initial data cleaning work for you. We were always glossing over how to collect and structure data that you might want to use.

Now, I could have gone and just played with another nice dataset that already existed (Keras has a nice list here or you can look at datasets on Kaggle) to save time and energy. As Big Data Borat says

In Data Science, 80% of time spent prepare data, 20% of time spent complain about need for prepare data.

— Big Data Borat (@BigDataBorat) February 27, 2013

I wanted to make sure I was using the proper data augmentation techniques and, as I added data and played around with optimization, that I understood every piece, which…