AISaturdayLagos: Recap on Week 2
We held week 2 of AI Saturdays Lagos on the 13th of January, 2018. People came in masses — 1, 10, 20, … 80.
Here’s a quick recap written by Azeez Oluwafemi and I.
Table of Content
Practical Deep Learning
We were able to ensure a large percent of the class had set up their google cloud instance while the remaining shared with them.
We watched videos by Jeremy Howard on making neural networks uncool again. It was a practical session so we had to do a code walk-through. The first code session involved trying out an binary image classification challenge from kaggle called “Cats and Dogs” without knowing the details of the architecture yet. So a resnet35 architecture was used at high level with details implemented in fast.ai framework which was also implemented in Pytorch. Students were challenged with the suggestion on how they could tweek the code to build their own binary classification project.
Image Classification
We also watched the Lecture 1 video of CS231n taught by Prof. FeiFei Li, Justin Johnson and Serena Yeung. For the first lecture, Prof FeiFei gave a brief history of Computer Vision and an overview of CS231n.
Evolution Big Bang : the number of animal specie exploded from a few of them to hundreds of thousand because the first animal developed eyes. — Andrew Parker
She talked about the challenges of vision and why it is important for computers to develop vision. She related this to building algorithms.
Lecture 2 was taught by Justin Johnson, a soft introduction to Image Classification, the motivation, example and challenges. He further discussed algorithms that can be used to solve the problem of Image Classification — Nearest Neighbor and Linear Classification.
Deep Learning Theory Discussion
Due to the nature of the complexity of the papers treated and in an effort to make the information accessible to enthusiast, we decided to make it a discussion class. That way people are forced to explain what they read and not just throw equations around. Those who wanted to know the details could however study and help provide interesting and reasonable intuitions for other students to understand. It ended up becoming one of the most interesting classes.
A lot of new concepts were introduced this week which we would be building upon in coming weeks.
Lagos chapter wouldn’t have happened without my fellow ambassador Azeez Oluwafemi, our Partners FB Dev Circle Lagos, Vesper.ng and Intel.
A big Thanks to Nurture.AI for this amazing opportunity.
Also read how AI Saturdays is Bringing the World Together with AI
See you next week 😎.
Links to Resources
Practical deep learning for coders: http://forums.fast.ai/t/welcome-to-part-1-v2/5787
Deep learning Theories: https://stats385.github.io/readings 3
Convolutional Neural Networks: http://cs231n.github.io/
A friendly introduction to Convolutional Neural Networks and Image Recognition: https://www.youtube.com/watch?v=2-Ol7ZB0MmU
