A pet that goes missing is among many people's worst fears: a moment ofdistraction is enough for a dog or a cat wandering off from home. Some measureshelp matching lost animals to their owners; but automated visual recognition isone that - although convenient, highly available, and low-cost - issurprisingly overlooked. In this paper, we inaugurate that promising avenue bypursuing face recognition for dogs. We contrast four ready-to-use human facialrecognizers (EigenFaces, FisherFaces, LBPH, and a Sparse method) to twooriginal solutions based upon convolutional neural networks: BARK (inspired inarchitecture-optimized networks employed for human facial recognition) and WOOF(based upon off-the-shelf OverFeat features). Human facial recognizers performpoorly for dogs (up to 60.5% accuracy), showing that dog facial recognition isnot a trivial extension of human facial recognition. The convolutional networksolutions work much better, with BARK attaining up to 81.1% accuracy, and WOOF,89.4%. The tests were conducted in two datasets: Flickr-dog, with 42 dogs oftwo breeds (pugs and huskies); and Snoopybook, with 18 mongrel dogs.
展开▼
