@luchaoqi.bsky.social
Ph.D. student advised by @ronisen.bsky.social.
Interested in generative AI.
Opinions are my own.
https://luchaoqi.com/
Interested in generative AI.
Opinions are my own.
https://luchaoqi.com/
4/N
📌 Huge thanks to our incredible team for making this work possible!
Jiaye Wu, Bang Gong, Annie N. Wang, David W. Jacobs, @ronisen.bsky.social
📌 Huge thanks to our incredible team for making this work possible!
Jiaye Wu, Bang Gong, Annie N. Wang, David W. Jacobs, @ronisen.bsky.social
November 27, 2024 at 3:51 PM
4/N
📌 Huge thanks to our incredible team for making this work possible!
Jiaye Wu, Bang Gong, Annie N. Wang, David W. Jacobs, @ronisen.bsky.social
📌 Huge thanks to our incredible team for making this work possible!
Jiaye Wu, Bang Gong, Annie N. Wang, David W. Jacobs, @ronisen.bsky.social
3/N
📌 Video-reaging: Our approach extends seamlessly to video re-aging using face-swapping techniques.
🚀 Explore our teaser video and project website for more exciting results!
📌 Video-reaging: Our approach extends seamlessly to video re-aging using face-swapping techniques.
🚀 Explore our teaser video and project website for more exciting results!
November 27, 2024 at 3:51 PM
3/N
📌 Video-reaging: Our approach extends seamlessly to video re-aging using face-swapping techniques.
🚀 Explore our teaser video and project website for more exciting results!
📌 Video-reaging: Our approach extends seamlessly to video re-aging using face-swapping techniques.
🚀 Explore our teaser video and project website for more exciting results!
2/N
📌 Solution: We personalize a pre-trained aging encoder (SAM) using ~50 personal images spanning 20-40 years.
🔧 Our age adapter network refines the W+ latent code from SAM, optimizing it for the most personalized representation within StyleGAN2’s latent space.
📌 Solution: We personalize a pre-trained aging encoder (SAM) using ~50 personal images spanning 20-40 years.
🔧 Our age adapter network refines the W+ latent code from SAM, optimizing it for the most personalized representation within StyleGAN2’s latent space.
November 27, 2024 at 3:51 PM
2/N
📌 Solution: We personalize a pre-trained aging encoder (SAM) using ~50 personal images spanning 20-40 years.
🔧 Our age adapter network refines the W+ latent code from SAM, optimizing it for the most personalized representation within StyleGAN2’s latent space.
📌 Solution: We personalize a pre-trained aging encoder (SAM) using ~50 personal images spanning 20-40 years.
🔧 Our age adapter network refines the W+ latent code from SAM, optimizing it for the most personalized representation within StyleGAN2’s latent space.
1/N
🤔 How can we simulate a person’s aging across their entire lifespan—for instance, Al Pacino transforming from age 0 to 100?
😇 By personalizing a pre-trained global aging prior with just 50 selfies, we achieve high-fidelity re-aging results while preserving identity seamlessly!
🤔 How can we simulate a person’s aging across their entire lifespan—for instance, Al Pacino transforming from age 0 to 100?
😇 By personalizing a pre-trained global aging prior with just 50 selfies, we achieve high-fidelity re-aging results while preserving identity seamlessly!
November 27, 2024 at 3:51 PM
1/N
🤔 How can we simulate a person’s aging across their entire lifespan—for instance, Al Pacino transforming from age 0 to 100?
😇 By personalizing a pre-trained global aging prior with just 50 selfies, we achieve high-fidelity re-aging results while preserving identity seamlessly!
🤔 How can we simulate a person’s aging across their entire lifespan—for instance, Al Pacino transforming from age 0 to 100?
😇 By personalizing a pre-trained global aging prior with just 50 selfies, we achieve high-fidelity re-aging results while preserving identity seamlessly!