This is the repository of my capstone project for the Udacity Machine Learning engineer nanodegree. The code was developed in the dog_app jupyter notebook and the models was trained in a workspace provided by Udacity.
The main goal is to build an algorithm that takes an image path and if it’s a human image, it will predict the resembling dog breed. If there is a dog in the image, it will predict its breed.
-
- Exploratory data analysis.
-
- Built a face detector with a haar cascade classifier.
-
- Built a dog detector with a VGG-16 pre-trained model.
-
- Create a CNN model from scratch, train it and test it.
-
- Use transfer learning with VGG-16 model. Train and test it.
-
- Built a final algorithm whith the human detector, dog detector and the last transfer learning model.