AI Virtual Try On
The ability to visualize yourself wearing the clothes you’re looking at online is a technical barrier that few, if any, retailers have managed to overcome. However virtual try on services are moving rapidly forward, and new tools and models using AI, rather than 3D tech, are bringing us closer to this vision.
This project by the Chinese Kolors Team from Kuaishou allows users to upload an image of themselves, along with a product image, and then uses advanced image recognition and deep learning to place the garment on the person in the picture. The results vary and are far from perfect in all cases. But overall the model is pretty impressive.
The team behind the Kolors project, is a specialized R&D group at Kuaishou (also known as Kwai in international markets), a leading Chinese social media and video-sharing platform. The Kolors team has developed an advanced text-to-image generation model named Kolors, which is based on latent diffusion techniques. The model is trained on billions of text-image pairs and supports both Chinese and English inputs.
The Kolors model is open-sourced under the Apache-2.0 license, allowing for both academic and commercial use, provided that the necessary registrations are completed for commercial purposes. The team has also made the model accessible on Hugging Face.
You can test the Kolors Virtual Try On yourself for free on Hugging Face. If you want to use their API or use the output for commercial purpose, visit VirtualTryOn.net.
