Build a pipeline to process real-world, user-supplied images.
Given an image of a dog, the algorithm will identify an estimate of the canine’s breed.
If supplied an image of a human face, the code will identify the resembling dog breed.
Assess the Human Face Detector The submission returns the percentage of the first 100 images in the dog and human face datasets that include a detected, human face.
Use a pre-trained VGG16 Net to find the predicted class for a given image: dog_detector function returns True if a dog is detected in an image and False if not.
Assess the Dog Detector The submission returns the percentage of the first 100 images in the dog and human face datasets that include a detected dog.
CNN architecture of trained model attains at least 10% accuracy on the test set.
Model architecture that uses part of a pre-trained model with accuracy on the test set of 60% or greater.