Inspiration
For a majority of students that enter high school, college, and beyond, choosing an effective wardrobe that is readily available and accessible is an enormous challenge.
What it does
The neural network reads through a dataset, fashion mnist, in order to train and categorize clothing items into one of ten categories. We have also included a user interface to input user preferences and produce images of what a wardrobe would consist of for this user.
How I built it
We used tensorflow and keras to create a neural network model and used openCV to produce a series of images for a user based off of the type of style they prefer to wear on a day to day basis.
Challenges I ran into
Keras modeling initially produced a repetitive accuracy, around 10 percent. Many more trials yielded a maximum accuracy of 10 percent with a relu activation for our hidden layers. Additionally, in reshaping our numpy arrays, input dimensions were not compatible.
What I learned
We learned more about the various functions of tensorflow and keras in terms of configuring each neuron layer to produce the highest accuracy. We learned about the uses and functions of openCV to a much higher extent.
What's next for FashionAI
We plan to use our neural network to surf popular clothing brands and expand the classification database to include all possible clothes for purchase. From there, we would be able to touch up the GUI to produce multiple options and a smoother user interface for a wider range of styles and categories.
Log in or sign up for Devpost to join the conversation.