-
-
Users Can Add Their Hair Care Experience
-
Users Can Upload Photos of their Hair for Determining Their Curl Pattern
-
Users will Receive Product Recommendations Based On the Popular Choices Per Hair Type
Inspiration
Our team - Kendra and Kelly - both grew up with curly hair, and we've learned that curly-haired people have it hard! All the pictures of luscious, frizz-free curls can take a village and more to achieve, and we were tired of trying product after product only to find they made our hair greasy or didn't provide near the 200% moisture levels our hair needed. So, we decided to create a site to help others through their wavy and curly-haired journeys!
What it does
Our site allows established curly-haired people to input their hair type (Type 2, 3, or 4) and products that have worked for them to build a database of products that work best for them. Through our site, people who are looking for product recommendations can input their hair type and top products that have worked for others with their similar hair type will be outputted. Additionally, those who are new to understanding their curls can input a picture of their hair and our computer vision algorithm will detect the curl pattern they have to inform them.
How we built it
Our website was built with Bootstrap. The functionality was fleshed out with javascript to handle user data and PHP to process image inputs. For the computer vision algorithm, the training weights for the curl pattern detection were created using the YOLO algorithm on various images of each curl pattern. To integrate the curl pattern detection, we called a Python script, which was returned to PHP.
Challenges we ran into
The computer vision object detection was the most difficult piece of our project because it took most of the weekend to train. Additionally, when we trained images that included the face, the algorithm learned how to detect the top of the hair. We had to adjust the images so that only posterior hair shots were included in our training.
Accomplishments that we're proud of
We are proud of the features in our site. We believe that these features will allow fellow wavy- and curly-headed users to get more information on how to take care of their hair. Collecting the experience data of different users makes the recommendations more accurate.
What we learned
We learned more about web development and handling user data input. Additionally, we learned how to integrate Machine Learning into websites for full functionality. Also, it is ideal for our users to input images with shots from behind to ensure the most accuracy.
Overall skill-wise, we learned more about PHP, JavaScript, Web Dev, and YOLO.
What's next for PurelyCurly
In the future, we hope to create a more robust training set for the curl pattern detection. This training set will include more hair colors, hair lengths, picture qualities, and angles of hair photos to accommodate as many different users as possible. Furthermore, we hope to draw patterns in the ingredients we find users with similar hair types using. For example, if Coconut Oil is a recurring ingredient in the Type 2 Hair users lists, we could also recommend that the users look out for that ingredient and recommend products with that ingredient.
Log in or sign up for Devpost to join the conversation.