Solutions for the 2016 and 2017 assignments of the Stanford CS class on Convolutional Neural Networks for Visual Recognition (CS231n)
In this repo you can find my solutions for the CS231n course offered by Stanford on visual recognition. I thought it might be helpful for other fellow students to share them here.
The solutions are for both courses:
- CS231n Winter 2016
- CS231n Spring 2017
The Spring 2017 solutions use both deep learning frameworks: TensorFlow and PyTorch.
You can find the course notes here.
Here you can find the lecture videos for the CS231n Winter 2016 course:
And here are the lectures for the 2017 course:
You can download the original assignments, without the solutions, here:
- Assignment 1:
- Winter 2016: http://cs231n.github.io/assignments2016/assignment1
- Spring 2017: http://cs231n.github.io/assignments2017/assignment1
- Assignment 2:
- Winter 2016: http://cs231n.github.io/assignments2016/assignment2
- Spring 2017: http://cs231n.github.io/assignments2017/assignment2
- Assignment 3:
- Winter 2016: http://cs231n.github.io/assignments2016/assignment3
- Spring 2017: http://cs231n.github.io/assignments2017/assignment3
Useful things to know
- To download the datasets needed for the assignments, you will have to run the bash scripts in the /datasets folder of each assignment.
- When doing cross-validation in the first assignment, you might receive some overflow errors depending on what hyperparameters you select. It turns out, this is expected but do let me know if you find a solution to avoid this.
A big thank you to Andrej Karpathy, Justin Johnson, Serena Yeung and Fei-Fei Li for creating this amazing course. It blows my mind that we live in a time where everyone has free access to courses from one of the best universities in the world. Many thanks also to Clement Thorey for sharing his solutions as well. They have been very helpful in certain stages of the assignments.