I led the development of a unique technology for digital clothing try-ons.
Users can upload their photos and choose an outfit from the provided options to see how it looks on them. The outfit can be either real clothing or a 3D-rendered one.
We also conducted research and experiments to reduce inference costs and allow for more flexibility regarding the person's pose in the uploaded photo.
During this product development I:
- Developed and optimized classification and regression models for enhanced image and video analysis.
- Advanced 3D human pose detection using cutting-edge techniques to accurately simulate human movements.
- Enhanced the SMPL model through research, improving project functionality and realism.
- Fine-tuned Generative Adversarial Networks (GANs) to create high-quality digital fashion assets.
VFRs are effective in minimizing returns, increasing customer satisfaction, and promoting sustainability by reducing CO2 emissions from transportation.
There is potential for future research on using VFRs in 3D cloth simulation engines during pre-production and in designers' collections, which could significantly impact minimizing fashion waste and overproduction.
---
Vitola, Liseth Sierra (2024). Digital Innovation for Sustainable Fashion: Virtual Fitting Rooms as a Strategy to Minimize Waste and E-commerce Returns. Toronto Metropolitan University. Thesis. https://doi.org/10.32920/25164554.v1