Dress to Impress is a web application that allows users to virtually try on clothes without needing to visit a store in person. Developed during the HackUPC, a 36-hour hackathon in Barcelona, it was part of a challenge organised by Inditex. The original challenge was to find the most similar pieces of clothing to the ones you like, in an online store. We enhanced this with a feature enabling users to try on clothes virtually by uploading a picture of themselves, where a machine learning model overlays the garment onto the user.
- Virtual Try-On: Upload a photo and see how clothes look on you.
- Similarity Search: Find clothing items similar to your chosen pieces.
- User-Friendly Interface: Easy and intuitive navigation.
- Machine Learning Integration: Advanced ML models for accurate virtual try-ons.
Frontend was built using React.js + Typescript. Backend was developed mostly using Java Spring Boot, but we also utilised Python FastAPI that allows communication with the ML model. We chose ResNet image recognition model as its performance turned out to be better than VGG and Xception.
- HackUPC: For organising the hackathon where this project was conceived.
- Inditex: For providing the challenge and inspiration for this project.
Special thanks to the following contributors:
- Nikola Grujic (@nikolatechie)
- Vladan Cvjetkovic (@cvele0)
- Milan Jovanovic (@MJovanovic40)
- Lukas Pottner (@Potti1234)
Thank you for using Dress to Impress! Enjoy your virtual shopping experience.