Best Look

• 

  Landing Page

• 

  Taking picture

• 

  What it does

• 

  Output

Inspiration

We had a friend who couldn't decide on a haircut.

What it does

It scans your face and determines your face shape, after it provides different hairstyles that complement your outputted face shape

How we built it

First we fine tuned a CNN (VGG16) to classify face shapes. Then we created a backend api using flask, which we loaded our trained model onto. We pass images to this backend api through our frontend which was built in react.

Challenges we ran into

We had issues with overfitting when training the model. We had problems with many errors covering the front end. We also struggled to find a way to accurately capture the face, because it was returning the same face shape, because it scanned the entire image box and not just the face.

Accomplishments that we're proud of

We are proud of the accuracy of the model and of everything we have learned to accomplish this project. We are also proud of how the frontend looks.

What we learned

We learned how to fine tune pre-trained models. We also learned more about how CNN work and we learned how to use the react webcam library.

What's next for Best Look

Next for Best Look is to add facial hair and accessories. We would also like to eventually add a chatbot functionality that allows users to ask questions regarding hairstyles, etc.

Built With

• cnn
• flask
• pytorch
• react