The aim of the project is to build a pipeline to process real-world, user-supplied images. The algorithm will identify an estimate of the dog’s breed given an image. When the image is of a human, the algorithm will choose an estimate of a dog breed that resembles the human. If neither a dog or a human is detected, then an error message is output. Therefore, the models in place should be capable of detecting a dog or human in an image, classify the dog to its breed and classify a dog breed that the human resembles.
The dataset used here are provided by Udacity and is similar to Kaggle Dog Breek Classifier Dataset.The dataset is saved under S3 in Amazon Web Service.
This link [https://s3-us-west-1.amazonaws.com/udacity-aind/dog-project/lfw.zip] provides the images for human and can be used to train a human face detector.
This link [https://s3-us-west-1.amazonaws.com/udacity-aind/dog-project/dogImages.zip] provides the images for dog and can be used to training of dog related model .
Dog Breed Identification (https://www.kaggle.com/c/dog-breed-identification)