Fashion Similarity Study

Student: Ryan Magaya (2786968m@student.gla.ac.uk) | Supervisor: Prof. Craig Macdonald University of Glasgow - School of Computing Science

Purpose of This Study

Research Question: Which AI vision-language models perform best at finding similar fashion items when users upload real wardrobe photos (not professional product images)?

Why this matters: Existing fashion recommendation systems are trained on professional product photos. We're testing if they work equally well on the kinds of photos real users take (dress on hanger, on floor, worn by you). Your ratings will help determine which AI model is best for real-world fashion search applications.

Academic Goal: This research contributes to my dissertation on building AI-powered circular fashion marketplaces.

What You'll Do (5 min)

  1. Upload a dress photo
  2. View AI-recommended similar dresses
  3. Click which ones are "Similar" or "Not Similar"
  4. Repeat for 3-5 of your dress photos

Data Collected

  • Your dress photos and ratings
  • Anonymous session ID only
  • No names, emails, or personal info
  • Used for comparing 4 AI models

Your Rights

  • Voluntary and anonymous
  • Close browser anytime to withdraw
  • Request deletion: email session ID

By clicking "I Agree and Start":

  • You are 16+ years old
  • You have dress photos ready
  • You will NOT upload photos with faces
  • You consent to anonymous data collection

Questions? 2786968m@student.gla.ac.uk