Lesson 1 official resources and updates ✅

jeremy · October 25, 2018, 4:21am

Lesson 1 - Welcome and Image Recognition

This topic is for official updates and information regarding lesson 1. Only admins are able to reply to this thread, so please subscribe to topic notifications to ensure you don’t miss anything. You should also follow the general course update thread.

Note that this is a forum wiki thread, so you all can edit this post to add/change/organize info to help make it better! To edit, click on the little edit icon at the bottom of this post. Here’s a pic of what to look for:

Lesson resources

Course site, including setup guides for each platform
Course repo
fastai docs
fastai datasets
Notebooks:
- 00_notebook_tutorial.ipynb
- lesson1-pets.ipynb
Detailed lesson notes - thanks to @hiromi
Lesson notes - thanks to @PoonamV (wiki thread - please help contribute!)

Other resources

Thread on creating your own image dataset
What you need to do deep learning (fast.ai blog post including some basics on what GPUs are and why they’re needed)
Original Paper for Oxford-IIIT Pet Dataset
The Oxford-IIIT Pet Dataset
What the Regular Expressions in the notebook meant
Understanding Regular Expressions (12 minute video)
Visualize Regular Expressions
Interactive tutorial to learn Regular Expressions
Beginners Tutorial of Regular Expression
One-Cycle Policy Fitting paper
Visualizing and Understanding Convolutional Networks (paper)
References in the video and in course page

How to scrape images

Video Timeline

00:00:01 Welcome and tools presentation
00:05:26 About jeremy and the fastai course
00:10:23 Lesson 1 start
00:17:55 Looking at the data
00:26:29 ImageDataBunch
00:35:05 Training: resnet34
00:40:04 Overfitting and validation set
00:41:40 fit_one_cycle
00:48:38 fastai and students achievements
01:08:41 Visualizing results
01:14:40 Unfreezing, fine tuning and learning rates
01:26:54 Training: resnet50
01:35:06 fastai doc notebooks

Detailed video timeline - thanks to @melonkernel

Resources: Lesson 2 >>>

jeremy · October 25, 2018, 4:30am

The edited lesson video has been added to the top post of this thread now.

jeremy · October 26, 2018, 4:02am

You will find a new download_images.ipynb in your nbs folder after you git pull. It shows how to download a dataset for training a model of your own. Many thanks to @lesscomfortable for his help on building this.

Note that this requires the latest fastai, just released today. To get it, go to http://course-v3.fast.ai/ , click ‘returning to work’ in the sidebar, and follow the instructions. Generally, they will be somelike like:

Click New->Terminal in Jupyter, then type:
conda update conda
conda install -c fastai fastai

jeremy · October 26, 2018, 4:31am

@part1_v3 Note that we have consolidated the forums so there is now just one topic you have to follow to get all official course updates, and one topic per lesson for all official lesson updates. This topic you’re reading now is the topic for lesson 1. Please scroll to the top and read to see that resources available and updates. And here is the topic for the course as a whole:

For each, I suggest that you select Watching from the notification list at the bottom of each of these topics, as shown below. That way, you won’t miss any important announcements. (We won’t be making any further announcements over email.)

jeremy · October 28, 2018, 6:16pm

Note that after you conda update and git pull, you’ll find that there’s been one change to fastai - ConvLearner is now called create_cnn. You’ll need to make that change in any notebooks that you’ve created too.

jeremy · October 28, 2018, 9:42pm

There are now 2 sets of course notes kindly created by students - both added to the top post of this thread.