HaircutHelper

• 

  Main Page

• 

  Picture Taker Page

Inspiration

Though haircuts may seem relatively unimportant, they have the ability to greatly boost confidence when done just right. We want to provide the blueprint for the best possible haircut for people all around the world in order to allow them to be the best versions of themselves, both externally and internally.

What it does

HaircutHelper provides personalized haircut recommendations for users based on facial structure and displays previews for potential hairstyles. Users have the option to either take a real-time image using their camera or to upload an image already stored in their computer. HaircutHelper then takes this data and by utilizing the power of machine learning, it compares the image with countess others to tell the user their face shape; it then uses face shape and confidence level to intelligently recommend to the user a specialized haircut blueprint.

How we built it

We built our project using HTML and Python. We collaborated with one another through GitHub Desktop and VS Code. We used HTML & CSS for the front end development. On the backend, we used python and trained our machine learning KNN model with 2 datasets for face shapes with 4 categories. We used flask framework to connect our backend and frontend.

Challenges we ran into

We had to change ideas mid way, as we had to completely change our data due to changing ideas from an injury-related detector to a haircut recommendation app. We had to learn how to train a machine learning for the first time learn and furthermore, how to use open cv for the first time. We needed to find resources to help with such tasks and we constantly struggled to learn how to properly use KNN. In addition we had to deal with the libraries we were working with not working (i.e openai , tensorflow ).

Accomplishments that we're proud of

We are proud of ultimately creating a project which we can see others using in the future to decide upon haircuts and boost their confidence. We are proud of our personalized recommendation system, and how our model uses countless training to take a photo and turn it into a personalized haircut recommendation (using information such as confidence level and face shape). We are also proud of our User Interface and front-end, which we believe we spent a long time working on to make it user-friendly, bug free, and visually appealing. Ultimately, we take pride in persevering through many challenges we encountered and creating a project which pushed us out of our comfort zones.

What we learned

We learned a lot of new things with this project specifically how to train a machine learning model with images using OpenCV. We also further explored the openai API and debugged issues regarding DALL-E, which was a big learning process for us. We also learned valuable lessons including strong communication between front-end and back-end members for everything to work as intended. Overall, this hackathon was a valuable learning experience for us both as a team and as developers.

What's next for HaircutHelper

Our next steps would be to implement another major plan we had in mind for the site which we were not able to implement due to time constraints. This would be to use DALL-E and AI/Augmented Reality to take the photo of the user and to create a new photo which demonstrates how the user would look with the specified haircuts. We believe this feature would further improve our project's usability.

Built With

• css
• cv2
• flask
• html
• openai
• python