Just for the Christmas to have some fun, we tried to train a deep learning classifier using 25000 images of cats and dogs. To simplify using it, one of the colleagues added an interface to it. Then we started testing it with images of cats and dogs. The performance is reasonable but there are obvious places to improve. It currently does not support black and white. You can fool the model by uploading data that it has not seen. For example it is weakly trained on Husky, or if you upload a picture upside down, it would be confused.
Now the fun part! If you upload a picture of a person, it tells which visual features are dominant and if it were to classify it into a cat or a dog, which one it will be. Following the round form of heads, form of ears, nose, and hair, majority of human faces are classified as cats. However, there are human faces that are yet classified as dogs. If you are also curious whether your face is classified as a dog or cat, check out : catordog.machine2learn.com
ps. With respect to the privacy, images are only processed once and no image is stored.
ps2. Leave a comment, if you find faces that are classified as a dog.