Homework 5 – part II simple image classification pipeline

Computer Vision
Homework 5 – part II
Q1 – mandatory for UG students, due Sunday, May 6th, by 11:55pm
Q1 + Q2 – mandatory for graduate students, due Wednesday, May 9th, by 11:55pm
Q2 – extra-credit for UG students, due Wednesday, May 9th, by 11:55pm
Q3 – extra-credit for graduate students, due Wednesday, May 9th, by 11:55pm
Part II: Training and Testing
In the second part of this assignment you will train and test a simple image classification
pipeline based on the SVM classifier and a simple two-layer neural network classifier.
Provided files:
• Starter code is provided as a zip file on Moodle. Your file tree opens in the
hmwk5 folder
• Download the CIFAR-10 data set for Python here:
https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz and save the archive in
the hmwk5_2/datasets folder.
Working locally
Installing Anaconda: To work locally, we recommend using the free Anaconda
Python distribution, which provides an easy way for you to handle package
dependencies. Please be sure to download the Python 3 version, which currently
installs Python 3.6.
Anaconda Virtual environment: Once you have Anaconda installed, it makes sense
to create a virtual environment for this assignment. If you choose not to use a virtual
environment, it is up to you to make sure that all dependencies for the code are
installed globally on your machine. To set up a virtual environment, run (in a
terminal)
conda create -n hmwk5 python=3.6 anaconda
to create an environment called hmwk5. You can of course choose a different name.
Then, to activate and enter the environment, run
source activate hmwk5
To exit, you can simply close the window, or run
source deactivate hmwk5
Note that every time you want to work on the assignment, you should run source
activate hmwk5 (change to the name of your virtual env).
You may refer to this page for more detailed instructions on managing virtual
environments with Anaconda.
Python virtualenv: Alternatively, you may use python virtualenv for the project. To
set up a virtual environment, run the following:
cd hmwk5
sudo pip install virtualenv # This may already be installed
virtualenv -p python3 .env # Create a virtual environment
(python3)
# Note: you can also use "virtualenv .env" to use your default
python (please note we support 3.6)
source .env/bin/activate # Activate the virtual environment
pip install -r requirements.txt # Install dependencies
# Work on the assignment for a while ...
deactivate # Exit the virtual environment
Working remotely on Google Cloud (not supported by our staff)
Another option is to use Google Cloud for this assignment. Google Cloud is free to try
for one year. You will definitely have better CPU/GPU resources than you may have
locally. Please see the set-up tutorial from Stanford’s cs231n here for more details.
Download data:
Once you have the starter code (regardless of which method you choose above), you
will need to download the CIFAR-10 dataset. Go to
https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz and save the archive in
the “datasets” folder. Run the following from the hmwk5 directory:
cd hmwk5_2/datasets
./get_datasets.sh
Start your virtual environment:
After you setup your virtual environment, just remember to run
source activate hmwk5
Start IPython:
After you have the CIFAR-10 data, you should start the IPython notebook server
from the hmwk5 directory, with this command:
jupyter notebook
(See the Google Cloud Tutorial for any additional steps you may need to do for
setting this up, if you are working remotely).
Once your notebook server is running, point your web browser at
http://localhost:8888 to start using your notebooks. If everything worked correctly,
you should see a screen like this, showing all available IPython notebooks in the
current directory:
Click through one of the notebook files (.ipynb) to work on different parts of the
assignment. Open knn.ipynb first; the solution is provided.
What You Have to Do:
Q0: k-Nearest Neighbor classifier
The IPython Notebook knn.ipynb will walk you through implementing the kNN
classifier. The solution for knn is provided. It will help you familiarize yourself with
the pipeline, the initial setup part, splitting the training and the testing data, calling
specific class files, doing cross-validation at the end.
Q1: Training a Support Vector Machine (50 points) – mandatory for all students
The IPython Notebook svm.ipynb will walk you through implementing the SVM
classifier. Fill in the missing code parts and answer the questions which require a
written answer. At some points, you will need to fill in missing parts in the class
files: hmwk5_2/classifiers/linear_classifier.py and hmwk5_2/classifiers/linear_svm.py.
Q2: Two-Layer Neural Network (100 points) – mandatory for graduate students
and extra-credit for undergraduate students
The IPython Notebook two_layer_net.ipynb will walk you through the
implementation of a two-layer neural network classifier. Fill in the missing code
parts and answer the questions which require a written answer. At some points, you
will need to fill in missing parts in the class files: hmwk5_2/classifiers/neural_net.py.
Q3: Implement a Softmax classifier (50 points) - extra-credit for graduate
students
The IPython Notebook softmax.ipynb will walk you through implementing the
Softmax classifier. Fill in the missing code parts and answer the questions which
require a written answer. At some points, you will need to fill in missing parts in the
class files: hmwk5_2/classifiers/softmax.py.
Submitting the assignment:
Save each IPython notebook and each class file as a pdf file. Zip your entire file tree (the
hmwk5 folder), plus the pdf files and submit the archive through Moodle, as Hmwk5-
partII before the deadline specified at the top.