Our code may be good... but your dress code will be better! We thought that this would be a great chance to tackle a stylish struggle in a fun and exciting way.
What it does
It is a tool that can be used to help you find something new and refreshing to wear with those plain old jeans. After thorough fashion research into the latest trends and insight into colour theory, our web application can help you find just the right thing to ramp up your style.
How I built it
We used data scraping to search a chosen website for suitable complementary items of clothing. We determined the optimal colour of clothing using colour theory and matrix manipulation. Machine learning was used to classify clothing and extract information on the colours in an item owned by the user. These would be compared with other items online. Description and link were both obtained by sifting through the html using beatiful-soup.
Challenges I ran into
The biggest challenge that was faced was struggling to link the front-end and back-end. A lack of knowledge of flask proved to be very unfavourable in this situation. However, we managed to complete working code on either end.
Accomplishments that I'm proud of
The team-members managed to train and test a convolutional neural network to 92% accuracy on evaluation. This was better than any previous efforts and marked a personal achievement. The members hadn't been used to working with html or css and gained a lot of experience in producing a working front-end.
What I learned
We all learnt a lot about how the back-end and front-end in a working web app interact with each other even though we had no prior experience in this area. This challenge made the experience even more interesting for us.
What's next for AutoStyle
We would ideally like to train the CNN on hundreds of different clothing styles of specific items, increase the range of classified colour schemes and work on a more accessible interface to further improve both the matching process and the ease of use. We aim to help all those people out there (like us) that struggle with deciding what to wear!