Dog Breed Classification using Inception-ResNet-V2

Dogs are one of the most faithful and loyal animals in the world.They are also the favourite pets for most of the pet lovers.Many feel relieved from stress and tension when they spent time with their pet dogs.So these special creatures are spread into various breeds across the world.It is very much essential to distinguish the breeds at many occasions.With the advent of development of artificial intelligence the methods to classify such large scale of breeds had become easier.This paper proposes a transfer learning based pretrained deep CNN architecture for classification of 120 breeds.The proposed model was trained on Stanford dogs dataset and the model achieved a training accuracy of 95.03% and a validation accuracy of 92.92% after training.The model performance and robustness had been inferred after testing with test images from internet.The network predicted correct breeds with a test accuracy of 88.92%.This paper provides an optimal solution for fine grained dog breed classification.