/doggo-vs-kitter

Image recognition on the Assira CAPTCHA.

Primary LanguageJupyter Notebook

Doggo-or-Kitter

Computers have a hard time distinguishing dogs from cats. So much so that this task was widely used as a CAPTCHA, known as Assira (Animal Species Image Recognition for Restricting Access). Recently, research has shown that machine learning approaches can crack Assira with an 80% success rate [1], leading to its retirement.

This ongoing competition is used to benchmark the latest computer vision and deep learning approaches to this problem.
Data courtesy of Microsoft Research, with a total of 37,500 images of cats and dogs.

Setup

To quickly setup the opencv and tensorflow depencies in Anaconda3 (Python 3):

# OpenCV is a computer-vision library.
conda install -c https://conda.anaconda.org/menpo opencv3

# TensorFlow is Google Brain's second generation machine learning system.
conda install -c https://conda.anaconda.org/jjhelmus tensorflow

# Keras is a modular neural networks library.
pip install keras

References

  1. Golle, P. "Machine Learning Attacks against the Asirra Captcha." Proceedings of the Acm Conference on Computer and Communications Security. (2008): 535-542. Print.