Mori Liu built an app that photographs your clothes and creates a searchable digital closet. He tried OpenAI for image recognition but switched to GDINO and DenoBlahana for better accuracy. The app organizes outfits by category, brand, and tags.
Let's build an app that digitalizes your entire closet. Hi, my name is Mori Liu, and I shop way too much to the point where I can't even remember all the stuff I have. So rather than buying fewer clothes, I decided to build an app that turns my outfit photos into a digital closet. I first decided on the design system and mapped out the user flow. Then I used Replit to turn my hand-drawn wireframe into clickable designs and implement them on the Replit platform. After frontend is done, I start to build API endpoints and backend. For clothing segmentation and generation, I was inspired by this tweet that used OpenAI model for both. However, after a few rounds of evals, we decided to swap the segmentation model to GDINO for better better object detection and the generation model for DenoBlahana, which preserves more fidelity of the reference photo. Finally, I hooked up the endpoints to the frontend. Now I can upload my photos and create an entire digital closet from them. I can also organize by category, brand, and search through my clothes. As usual, come and closet if you want to try it out, and let me know what other features you want to see.
