In this project, I built a Convolutional Neural Networks (CNN) that classify 1 of the 133 dog breeds in the training dataset. If you supply a picture of your face or your friend's face, the app will also tell you the closet match of a breed.
This is the capstone project for my Data scientist nano degree. You can find more about the Nano degree here
If you want to train a model yourself, follow the instructions here. If on the other hand, you just want to run the app, follow the following instructions:
- Clone the repository and navigate to the downloaded folder.
git clone https://github.com/tuanchris/dog-project.git
cd dog-project
- Install required packages
- Linux or Mac:
conda create --name dog-project python=3.5
source activate dog-project
pip install -r requirements/requirements.txt
- NOTE: Some Mac users may need to install a different version of OpenCV
conda install --channel https://conda.anaconda.org/menpo opencv3
- Windows:
conda create --name dog-project python=3.5
activate dog-project
pip install -r requirements/requirements.txt
- Dog breed/human detection
python dog_app.py /path/to/image
This model was able to achieve 81% accuracy on the test set.
Here are some further improvements for the model
- Augment training data to prevent overfit
- Tune model parameters to improve accuracy
- Try other models