Lesson 1 - Introduction

Lesson Video:

About this course/study group:

• It's a hybrid
• Like Practical Deep Learning for Coders but many bonuses
• Spans 2-3 hours depending on who stays with 1-1.5 hour lectures

About Me:

• Been doing fastai for 1.5 years
• A few projects (some involving fastai, some not)
• University undergraduate, Software Design and Development with a minor in ES

Why did I make this?

• Make fastai accessable to Undergraduates at my school (University of West Florida)
• We both learn
• Techniques not mentioned in the course

Course Format:

Below you will find links to all the lessons and their notebooks in this course. To open up and use them, simply click the "Open in Colab" button at the top of each

Vision Block:

• Lesson 1:
  • Introduction
  • Pets and the High Level API
  • Image CLassification with a Custom Dataset
• Lesson 2:
  • Low-Level API with MNIST and Building Models
  • Stochastic Gradient Descent
  • Deployment
• Lesson 3:
  • Multi-Label Classification
  • Recognizing Unknown Images (or the Unknown Label problem)
  • Cross Validation and Ensembling
  • The Internal API of fastai
• Lesson 4:
  • Image Segmentation
  • ImageWoof and Exploring SOTA in fastai
  • Debugging with the DataBlock
• Lesson 5:
  • Style Transfer
  • Deployment Continued
  • EfficientNet and Custom Weights
• Lesson 6:
  • Keypoint Regression
  • Hybridizing Models
  • Object Detection
  • Multimodal Learning with Bengali.AI
• Lesson 7:
  • Super Resolution
  • Siamese
  • Audio

Tabular Block:

• Lesson 1:
  • Introduction to Pandas
  • The Tabular API with Binary Classification
• Lesson 2:
  • Regression and Permutation Importance
  • Ensembling with Other Libraries
  • Bayesian Optimization
• Lesson 3:
  • SHAP and fastinference
  • ClassConfusion and fastinference
  • Doing Research
• Lesson 4:
  • TabNet
  • Time Series

Requirements:

• A Google account to use Google Colaboratory (what I am in right now)
• A Paperspace account for Natural Language Processing

Communication

• Key to making study groups work!
• We have a dedicated mega-thread on the fastai forums here for us to use!
• No question is too stupid!

Environment setup

Google Colaboratory - Free GPU's

Go to here to start using it.

File -> Open notebook

Select from GitHub and use the following URL: muellerzr/A-walk-with-fastai

Choose open in a new tab

And now your screen looks similar to below! (With what I am using instead)

Changing our processer to a GPU

• 'Runtime'
• 'Change runtime type'
• 'Runtime type' should be 'Python 3'
• 'Hardware accelerator' should be 'GPU'

Initial Run Warning (For pulling from GitHub)

If you get a pop up saying 'Warning: This notebook was not authored by Google,' check 'Reset all runtimes before running' and click 'Run Anyway'

Saving your work

For now this will just save the notebook on their cloud, we will look next week how to utilize your Google Drives!