Post

πŸ‘šπŸ‘š ViViD Diffusion Virtual Try-ON πŸ‘šπŸ‘š

πŸ‘‰Alibaba announces ViViD, a novel framework employing powerful diffusion models to tackle the virtual try-on task.

Code announced, not released yet😒

𝐇𝐒𝐠𝐑π₯𝐒𝐠𝐑𝐭𝐬:

  • βœ…Novel architecture to address the video VTON
  • βœ…Diffusion models to synthesize HQ try-on videos
  • βœ…Pose + temporal modules for temporal consistency
  • βœ…New attention feats. fusion mechanism for garments
  • βœ…Multi-category dataset: 9,700 pairs of HQ garment-clips

πŸ‘‰Discussion https://t.me/s/AI_DeepLearning

πŸ§™Paper Authors: Zixun Fang1,2βˆ— Wei Zhai1† Aimin Su2 Hongliang Song2 Kai Zhu2 Mao Wang2 Yu Chen2† Zhiheng Liu1 Yang Cao1 Zheng-Jun Zha1 1University of Science and Technology of China 2Alibaba Group

Translate to Korean

πŸ‘‰ Alibaba λŠ” 가상 μ²΄ν—˜ μž‘μ—…μ„ μ²˜λ¦¬ν•˜κΈ° μœ„ν•΄ κ°•λ ₯ν•œ ν™•μ‚° λͺ¨λΈμ„ μ‚¬μš©ν•˜λŠ” μƒˆλ‘œμš΄ ν”„λ ˆμž„μ›Œν¬μΈ ViViDλ₯Ό λ°œν‘œν–ˆμŠ΅λ‹ˆλ‹€.

μ½”λ“œ λ°œν‘œ, μ•„μ§πŸ˜’ κ³΅κ°œλ˜μ§€ μ•ŠμŒ

ν•˜μ΄λΌμ΄νŠΈ:

  • βœ…λΉ„λ””μ˜€ VTON을 λ‹€λ£¨λŠ” μƒˆλ‘œμš΄ μ•„ν‚€ν…μ²˜
  • βœ…HQ μ‹œμ°© λΉ„λ””μ˜€λ₯Ό ν•©μ„±ν•˜κΈ° μœ„ν•œ ν™•μ‚° λͺ¨λΈ
  • βœ…μ‹œκ°„μ  일관성을 μœ„ν•œ 포즈 + μ‹œκ°„μ  λͺ¨λ“ˆ
  • βœ…μƒˆλ‘œμš΄ μ£Όλͺ© μœ„μ—…. μ˜λ³΅μ„ μœ„ν•œ μœ΅ν•© 기계μž₯치
  • βœ…λ‹€μ€‘ λ²”μ£Ό 데이터 μ„ΈνŠΈ: 9,700케레의 HQ 의λ₯˜ 클립
This post is licensed under CC BY 4.0 by the author.