Deep Learning Tutorial

Use this link to open the stubs notebook in Google CoLab:
http://colab.research.google.com/github/tolaw/dl-tutorial/blob/master/stubs.ipynb

The notebook used in the tutorial:
http://colab.research.google.com/github/tolaw/dl-tutorial/blob/master/reference.ipynb

The presentation in the tutorial:
https://github.com/tolaw/dl-tutorial/raw/master/Deep_Learning_Workshop.pdf

Hebrew Letters Challenge

Download the dataset from:
https://github.com/tolaw/dl-tutorial/raw/master/hebrew/train.zip

Pre-process the data, build and train a model that predicts the correct letter
Your solution should include:

1. Your code as a Jupyter notebook
2. A python function predict(image_path) that receives a path to an unprocessed image, runs pre-processing, predicts the correct label and returns it as a number (0, 1 or 2)

The models will be tested on a separate test dataset
The submitter of the best solution will receive a prize

Please submit your solutions to [email protected] before March 26

Reference Links

NumPy

numpy.ndarray
numpy.ndarray.reshape

Slicing NumPy Arrays

Keras

Sequential Model

keras.datasets.Sequential
keras.datasets.Sequential.compile
keras.datasets.Sequential.fit
keras.datasets.Sequential.evaluate

Keras Optimizers
Keras Loss Functions

Layers

keras.layers.Dense
keras.layers.Dropout
keras.layers.Flatten
keras.layers.convolutional.Conv2D
keras.layers.convolutional.MaxPooling2D

Keras Weight Initialization
Keras Activation Functions

Visualization with MatPlotLib

PyPlot Tutorial

Tutorials on YouTube

Deep Learning
Python for beginners