Inspiration
The inspiration behind this project was personal struggle when deciding outfits for trips, occasions etc. Even though we have a lot of clothes, we end up getting something new because of 1. Lack of ideas on how to style it differently, 2. We forget what we already have, 3. It is easier to look for an outfit instead of creating one.
What it does
It addresses all these problems, by going through your gallery understanding your personal style and remembering what you already own. Then you have an option whether you want an outfit just from these pictures or if you are open to trying a few new things like adding a scarf on top of an old dress. You can also add in ideas if you have already. You also need to set context of why you want this outfit and then it will suggest one for you.
How we built it
We used google api to first connect to photos where for now there was an album but later we can use our picture album as identified by google and choose only me to define the clothes we already have. Then we take all the inputs add it in the prompt and send it along the album.
Challenges we ran into
To limit the scope for phase 1, to make api work and give access to app, to make it faster, to create decent responses and not to hallucinate.
Accomplishments that we're proud of
That it is giving relevant answers and for our personal gallery that we did not add in video due to privacy reasons, it was working pretty well.
What we learned
Google systems, how to connect how to use google LLMs etc.
What's next for Closet companion
Right now everything is being directly called. In the next phase, we will be creating embeddings to account for huge albums to make retrieval faster and also to get a better overall understanding. Then clustering can be tried too, to create informational embedding of the pictures and match with the response user submitted. Further, once user tags the faces identified by google in the pictures, it can also give information like what we wore last time with a specific person. Then it can be linked to a shopping app as
Log in or sign up for Devpost to join the conversation.