‼️ See https://github.com/AI4LAM/fastai4GLAMS, future changes will live in this new repository. ‼️
A study group for v4 of the fastai Pracitcal Deep Learning for Coders course with a focus on applications in GLAM (galleries, libraries, archives, and museums)
Nicole Coleman is running a Elements of AI 4 GLAM study group which you might also want to check out as an alternative.
"Elements of AI is a free online course developed at the University of Helsinki that addresses the theory behind AI. The course is intended for people who want to learn what AI is, what can and cannot be done with AI. The study group provides a context for the material."
- 2020-09-01: Notebooks based on the course lesson notebooks can be found in lessons
- 2020-08-21: the course has now launched. See thread for discussion
- 2020-08-06: discussion thread for the study group now live on the forum: https://forums.fast.ai/t/glams-galleries-libraries-archives-and-museums-fastai-study-group/76064?u=danielvs
a study group:
- following v4 of the fastai Practical Deep Learning for Coders course with a focus on applications in GLAM settings
- focused on helping each other out
- tackling domain specific issues
- has a practical focus
This course will follow v4 of the fastai Practical Deep Learning for Coders which should be released sometine in July 🤞 is available at https://course.fast.ai/
- 🤔 Ethics
- 👀 CV (Computer vision)
- 📖 NLP (Natural Language Processing)
- 🗂 Tabular Data
- 🤖 Deployment of models
- 🐍 I have picked up ++ Python from doing previous versions of this course
There is a lot of interest in applying AI/ machine learning in GLAM settings with a range of potential applications being explored. There is also sometimes the perception that machine learning is very hard or can only be done by large tech companies. The fastai course aims to make deep learning (a branch of machine learning) acccesible while not hiding the important underlying theory. As a result I think this course is could be very useful for people working in/with GLAM instutions because:
- It will allow you to start using deep learning in a hands-on way, and you may be able to solve problems at work using machine learning
- Doing things yourself will give you a much better sense of what is possible and how things work
- If you collaborate with computer scientists, commercial companies etc. on ai projects, the course should help you to:
- a) be better able to communicate with them
- b) have a better sense of whether a vendor or commerical partner is overpromising
- c) be more aware of things about the GLAM setting that can pose challenges for machine-learning approaches
- 😃 It's fun!
There is no particular prerequisite for joining the study group. The fastai course assumes you have been coding for at least a year with the course using Python. There is some maths in the course but it is explained very clearly and I really wouldn't worry let math worries stop you from doing the course. If there are things you don't understand you can fill in the missing pieces as you go rather than wasting time learning things you might not need before getting started.
This study group will be primarily asynchronous with discussions taking place on the fastai forums. The reason this has been chosen over a slack group is that:
- there is already a ton of expertise on the fastai forums
- a slack channel is semi-private which means it's a bigger barrier to join and discussions get lost
- the forum discussions can be a resource that people might stumble upon through a google search in the future
The suggested approach will be:
- 🍿 watch the fastai course videos and work through the notebooks in your own time
- 🕸 use the study group forum thread to discuss the lessons, ask questions and share resources
- ✨ since data is so fundamental for deep-learning we may want to curate some labelled datasets which can be used to apply what is learned in the lessons.
- 📞 if there is interest we can set up a video call(s) at some point during the course
I will add notebooks to this repository based on materials from the fastai course to using GLAM data and will happily accept pull-requests for other notebooks.
When I did the course in 2019 I usually did one lesson per week. This comprises around ~2 hours to watch the video and then variable amount of time adapting the notebooks and trying to apply what is covered in the lessons doing follow up reading etc. The amount of time spent can be adjusted depending on your interest in each application and your own schedule.
- 🔔 The new version of the fast.ai course will launch on 21 August.
If you want to get an email reminder when the course starts add your email to the sign-up. - Join the thread on the fast.ai forums!