Why?

Because Getting Started is:

• Easy

• Less intimidating

• Less chance of getting stuck

How we will do it?

By:

• focusing on the process rather than just getting the output.

• starting with easier problems.

Prerequisite

• PyTorch: A 60 Minute Blitz, Writing Custom Datasets

Expectations from this series

Time: 5–6 Hours per challenge or more.

Coding Style: fastai coding style

• While coding, no function takes 10 seconds.

• Heavy use of %timeit%

Explanation Style: Based on Kaggle Notebook Commits

• Notebook Commit 1: Will focus on getting input and output

• Next N-1 Commits: Will add new features/improvements. These will be linked.

• Last Commit: Complete solution

This way one can see get a peek behind the finished product.

Abbreviations Used

X = training_data
y = training_label

ds = dataset
dl = dataloader

o = optimizer

Index

1. Digit Recognition
2. Facial Keypoints Detection - in progress