Multimodal / VLM2024
Virtual Try-On
See how a garment actually looks on you before buying, rendered by a vision-language model that reasons about fit.
VLMvision-language try-on
The problem
Shoppers cannot tell how a garment will actually sit on them from a product photo, and static size charts do not answer the question either.
What I built
I built a virtual try-on system on vision-language models: it reads a person image and a target garment, reasons about fit and drape, and renders the garment onto the person rather than pasting a flat overlay.
The outcome
A working try-on pipeline that treats the garment as something to understand, not just composite, giving shoppers a realistic preview before they buy.
Stack
- Python
- PyTorch
- VLM
- Diffusion
- ONNX
Working on something like this?
Tell me what you are trying to build. I will tell you honestly whether it is worth doing.