README.md

Introduction to AI and Deep Learning

For the Hour of Code / Hausner Code Fest, 6th to 8th graders.

Page easy url: http://bit.ly/code-fest

Session I: What is Deep Learning? The case of handwritten digit recognition (MNIST)

Neural Networks

Perceptron model, Feed-forward Neural Networks

Tensorflow playground: play with layers, features, and neurons.

http://playground.tensorflow.org/

MNIST online demos

Handwritten digit recognition demo: https://transcranial.github.io/keras-js/#/mnist-cnn

See the learning and optimization in action: http://cs.stanford.edu/people/karpathy/convnetjs/

Simple Neural Network with Keras and Python

(Notebook) https://github.com/ea167/code-fest/blob/master/notebooks/1_Simple-DNN-for-MNIST.ipynb

2 basic layers

Convolutional Neural Network with Keras and Python

(Notebook) https://github.com/ea167/code-fest/blob/master/notebooks/2_Convolutional_neural_networks-for-MNIST.ipynb

5 layers: 2 convolutional + 1 to group ("max pooling") + 2 layers

Part II: more advanced techniques - Style transfer

Image recognition ("ConvNet" or "CNN")

Demo Inception v3: https://transcranial.github.io/keras-js/#/inception-v3

If you feel doubtful, try these URLs as images :
- https://static.boredpanda.com/blog/wp-content/uploads/2017/09/funny-dog-bios-92-59ad3f9cd01a7__700.jpg
- https://static.boredpanda.com/blog/wp-content/uploads/2017/09/funny-dog-bios-70-59ad0f4a8264b__700.jpg
- https://static.boredpanda.com/blog/wp-content/uploads/2014/06/cute-bunnies-tongues-13.jpg
- https://images-eu.ssl-images-amazon.com/images/I/61qzr5q6l5L._SL1500_.jpg

Other pre-trained Networks: ResNet, SqueezeNet, VGG

Style transfer demo

Style Transfer: https://tenso.rs/demos/fast-neural-style/

Style transfer with Keras/Tensorflow and Python

(Notebook) https://github.com/ea167/code-fest/blob/master/notebooks/3_Artistic_style_transfer_with_a_repurposed_VGG_Net_16.ipynb

How to do it yourself at home

Install Docker (PC or Mac): https://www.docker.com/community-edition

Launch Docker, and download my image for AI|DeepLearning: docker pull ea167/jupyker-cpu

Then follow instructions here to launch: https://hub.docker.com/r/ea167/jupyker-cpu/~/dockerfile/

Or better if you have a PC with Nvidia GPU:

docker pull ea167/jupyker

and https://hub.docker.com/r/ea167/jupyker/~/dockerfile/

More Reading

http://keras.io

https://harishnarayanan.org/writing/artistic-style-transfer/ | Many images are borrowed from this great site.

http://neuralnetworksanddeeplearning.com

http://www.deeplearningbook.org/

Glossary

http://www.wildml.com/deep-learning-glossary/