What is the application?
It specifies the breed of dog by pictures using your device's camera or image gallery. And as can be recognized by the human photos which of dogs he most resembles. And just for fun, the application can guess which dog you or your friends could be.
How does it works?
Picture is input to the expert system based on neural networks and its output formed hypothesis what breed of dog is shown in this photo. The properties of neural networks such that even a photo on which there is no dog will be associated to varying degrees with the familiar dog breeds.
What is recognition accuracy?
The system trained to recognize 48 breeds of dogs by 40 thousand photos. In the last version of the application accuracy of breeds recognition was 84% in the 40 thousands of photos.
Experiments with neural networks are continuing, so the number of recognized breeds will be increased, and Recognition quality will improve in new releases.
Goals For Future.
It will be added to supplement the training set of dog photos your examples and thus continuously expand the number of dog breeds and recognition accuracy. The purpose of the project to create an expert system able to classify all known breeds of dogs from photos.