LookMate – Your virtual fashion companion.

About the Project

Inspiration

The idea behind the Virtual Try-On (VTON) app was inspired by the need for a more interactive and personalized shopping experience online. With the rise of e-commerce, shoppers often struggle to visualize how clothing items will look on them, leading to uncertainty and frequent returns. Our goal was to create a tool that allows users to virtually try on outfits, helping them make confident purchasing decisions.

What We Learned

Developing this app taught us a lot about integrating AI and machine learning into a web application. We explored advanced image processing techniques using a pretrained VTON model and gained valuable experience with cloud deployment, ensuring smooth performance. Additionally, working with APIs like Google Gemini and Hugging Face helped us understand the complexities of real-time data processing and chatbot interactions.

How We Built the Project

We built the VTON app using Streamlit for the frontend, creating an intuitive interface where users can upload body and garment images or choose from pre-loaded samples. A pretrained VTON model performs the virtual try-on, generating realistic image outputs. The app also features a chatbot powered by the Google Gemini Generative AI API to offer personalized outfit recommendations based on user preferences. We used the Hugging Face API to manage image processing, ensuring responsive try-on experiences, and deployed the app on Vultr cloud for scalability and reliability.

Challenges We Faced

One of the primary challenges was integrating the VTON model into a web app format without compromising performance. Achieving seamless real-time processing required efficient management of API calls and cloud resources. We also faced some hurdles with API rate limits and ensuring data security, especially when managing user-uploaded images. These challenges taught us the importance of optimizing resource usage and enhancing user privacy.