Project Voight-Kampff
Originally found HERE.
Beating Google's reCaptcha using AWS Rekognition. Part of project Touch-Captcha (두 터치). I did this because I cannot promote a better Captcha without first beating the industry standard.
Nothing special here. Credit goes to the ML researchers who developed the image classification technologies readily available today, either via the Google Vision API or AWS Rekognition.
Voight-Kampff comes from the movie Blade Runner (1982). It is the test used by Blade Runners to tell a Replicant(synthetic human/android) from a human being.
I am doing this because: I like research and I want to get a PhD in Machine Thinking (the inverse of Machine Learning). Your contributions will help me focus solely on this work with minimal distractions from the outside world.
You will need:
- Google GCP account (The virtual machines I use are hosted in GCP)
- AWS account (Proxies)
Pull with
curl "https://raw.githubusercontent.com/pirates-of-silicon-hills/test/master/setup.sh" --output setup.sh
chmod u+r+x setup.sh
./setup.sh
Past Puzzles
Every puzzle you see in my demonstration videos has been saved as an image with a unique name as identifier. You can download all the images here: https://drive.google.com/open?id=18b0HxyOsLP6AZMpF1-DNITrGvFBGkYND