This project uses a python script to download images from http://image-net.org/ with a catch to determine whether the link is valid or not.
The images are then downloaded, and from there a 6 layer neural network is created with tensorflow. The model then runs over 100 epochs to determine the accuracy to 96.42%.