kaggle-cifar10-extract

Python script for extraction of pixel data from CIFAR10 images. This will convert the images into an easily-accessible CSV file with the format:

"pxr0","pxg0","pxb0","pxr1","pxg1","pxb1","pxr2","pxg2","pxb2","pxr3","pxg3","pxb3",...,"class"
59,62,63,16,20,20,25,24,21,33,25,17,50,32,21,71,48,29,97,69,40,115,82,49,137,...,"horse"
...

Where pxr0 is red value for pixel 0, pxg2 is green value for pixel 2, etc.

Instructions

Download train.7z from the Kaggle site and extract the train folder to this repo directory. Also make sure the file trainLabels.csv is in the same directory as the repo. Then run:

python getpx.py > train.csv

Dependencies

This script depends on PIL (Python Imaging Library). You can probably install this by running this command (if you have Pip):

pip install PIL