[Study group] Lesson 1 - Image Recognition
Details
PLEASE READ THIS IN ITS ENTIRETY. There are prerequisite tasks to complete in order to benefit from attendance (e.g. lecture videos/homework).
The study group will be going through Lesson 1 of Fast AI Part 1. This will be open to newcomers, as last week's meetup was on short notice. Notes from the introductory meeting can be found here: https://github.com/ClevelandAIGroup/fastai-study-group/blob/master/Notes-Session0.md
Please note that we have no official connection with fast.ai, we only use their materials.
What is this workshop series about?
Over the next seven weeks, we will be following the 2018 version of the fast.ai course, Practical Deep Learning for Coders, Part 1. This course was originally delivered late last year at the University of San Francisco by Jeremy Howard (Kaggle's #1 competitor 2 years running). The course lectures and other materials have been kindly made available free online by fast.ai in MOOC form.
How are these workshops going to work?
Each session we will use that week’s fast.ai lectures and course materials as a basis for discussion and learning. We will invite everyone to contribute their insights and questions about it, work through important aspects of the notebooks etc.
What would my role be as a participant in the workshop?
We hope that everyone will be able to contribute while they are learning. Note that we do not formally “teach" deep learning in these sessions, but rather provide a participation-oriented environment for everyone to learn together.
What kind of time commitment is involved?
Prior to each workshop, we would request that you at watch the lecture for that week (typically these are around 2 hours long; shorter if you play them back at 1.25x speed or faster!). Work through homework and optionally the supplementary course materials.
You will need 10hrs/week on average (depends on your baseline and experience with python, numpy, pandas, pytorch and deep learning) to complete the course.
Online courses platforms completion rates are pretty low. To help keep people accountable, we suggest to collect a deposit from students. You choose the amount. At the end of the program you will get all the money back. If you dropout without completing the program we count the donation as sponsorship for future CAIG meetings.
What tools does the course use?
The fast.ai course is based around Python 3.6, so familiarity with numpy and pandas is required. For the deep learning component, fast.ai supplies its own package (fastai) which is built on top of PyTorch, a python package for tensor computation and deep learning.
I’m in. What do I need to do before the session?
- Sign up!
- Watch the first video at http://course.fast.ai/lessons/lesson1.html
- Do homework assignments:
- Setup environment on crestle/aws/paperspace/gcp/google collab/your computer/... (see Intro session notes for details)
- Get comfortable with Jupyter and all other tools
- Run week1 code and understand it…Play with code to understand it
- Try different learning rates, epochs while running code
- Run code on different set of images (download from google images for example) and share results
- Feel free to explore week2 notebook
- Take a look at the supplementary materials at http://forums.fast.ai/t/wiki-lesson-1/9398 and/or these excellent student notes: https://medium.com/@hiromi_suenaga/deep-learning-2-part-1-lesson-1-602f73869197
- Prepare any questions you may have and insights that you would like to share with the group.
Here are some more resources to help you to get started:
Course home page: http://course.fast.ai/index.html
First lecture: http://course.fast.ai/lessons/lesson1.html
Week 1 wiki: http://forums.fast.ai/t/wiki-lesson-1/9398
You might also find these excellent student notes useful: https://medium.com/@hiromi_suenaga/deep-learning-2-part-1-lesson-1-602f73869197
Course forum: http://forums.fast.ai/
Course notebooks on Github: https://github.com/fastai/fastai/tree/master/courses/dl1