I participate in a 10-week course of Siraj Raval(Make Money with Machine Learning). 500 students. Slack was used as a forum. Now discord is used. The focus is on Deployment. I want to use fast.ai in that course. Is anybody here participating? I would like to share information from the course so that we can find how to combine fast.ai and siraj’s deployment. Siraj’s goal is to spread AI to the world so this is in line with that.
Week 1:
Deploy Keras Model with Flask as Web App in 10 Minutes
Transfer learning image classifier
https://www.tensorflow.org/js/tutorials/transfer/image_classification
1 Like
All the live sessions with Siraj:
W1 https://www.youtube.com/watch?v=lv4Jv1QiSMc
W2 https://www.youtube.com/watch?v=UUYPENQPOVA
W3 https://www.youtube.com/watch?v=0-don-KKyUM
W4 https://www.youtube.com/watch?v=-nPjVoq5mdE
W5 https://www.youtube.com/watch?v=X37wHVwYN1U Here he is talking about fastai solutions!
W6 https://www.youtube.com/watch?v=ImaMmrbcWq0
W7 https://www.youtube.com/watch?v=1BE6aozvfZw
W8 https://www.youtube.com/watch?v=WeCdIs4EiUk
W9
I’m curious how his course turned out for you since there’s quite a bit of backlash against him about it in the media. While I’m sometimes inclined to believe it could be due to a vocal minority of ultra pissed off customers.
It turned out good. He shows how to put together different deployment parts. If you have a class of 500 (it became much more) there will always be disappointments. To easy or to complex etc etc. Yes, you are right. Those that was disappointed complained a lot. I don’t know how many but all that wanted a refund in the beginning did get it. I get a lot of stuff to combine so I can deploy fastai better.
1 Like
I took a closer look at some of his live coding sessions and I find that his coding style is very similar to mine, albeit moee refined. His approach is to search github to make sure he’s not reinventing the wheel. In practice, as a professional programmer, that’s what I tend to do. I download existing code, see if it works, and improve on it or rewrite it from scratch using the old code as a guide.
This might not work as well for fastai since it’s not as popular as tensorflow, keras, sklearn, etc and thus will have less github repos available.
I also liked how he took an idea of an AI startup from the audience and deployed it in minutes on the live coding session. Deployment has always been a headache for me.
I can definitely see myself taking his courses in the future… Maybe try to do all of his projects in Fastai as welll if i have the time.
Btw, i think plagiarism from github is less of a problem if you’re building an app as opposed to teaching a course. Nobody cares if you’re working for Facebook and you steal open source code from Github to incorporate into its latest algorithms.
3 Likes