Image classification environment set-up

Using this fork will get you up and running in a codespace in just a few minutes.

1. Get a Kaggle account

The dataset is hosted on Kaggle. To download it you need a free account. It's easy to set up via the following link:

Login or Register | Kaggle

After setting up your account, you will need to verify with a phone number to use the Kaggle API. You can do this from the settings tab in the menu revealed by clicking your profile picture at the top right of any Kaggle page.

Once you are registered and logged in:

• Go to the Dogs vs Cats competition page
• Go to the 'Data' tab
• Scroll down and click 'Join competition'

2. Generate a Kaggle API key

Now, you need to generate and save an API access token so that you can download the dataset from within a codespace:

• From the kaggle homepage, click on your profile picture in the upper right
• Select 'Settings'
• Scroll down, under API, click 'Create New Token'
• Click 'Continue'
• Save the key file on you local machine

The contents of the file should look like this:

{"username":"your-user-name","key":"a-bunch-of-letters-and-numbers"}

3. Start a codespace

Once you have your Kaggle username and API key, fork this repository and start a Codespace as you normally would.

4. Add your Kaggle API key

From your codespace, open the .env file and add your Kaggle username and key from the kaggle.json file you downloaded to your computer from the Kaggle site earlier.

export KAGGLE_USERNAME=your-user-name
export KAGGLE_KEY=a-bunch-of-letters-and-numbers

5. Download and prepare the data

I have placed a series of commands to download and organize the images in a shell script for you. To use the script, run the following command in your codespace terminal:

./get_data.sh

The script downloads the data from Kaggle, decompresses it and moves the training cats and dogs into separate directories.

Easy! Now we can work with the data in a Jupyter notebook. Open src/mvp.ipynb to get started.