📍Hyper-Dense Landmarks at 150FPS📍

👉#Microsoft unveils the SOTA in dense landmarking + #3D reconstruction. MAGIC.

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Accurate 10× as many landmarks as usual
✅Synthetic data, perfect annotations
✅NO appearance, light, diff-rendering
✅#3D @150+FPS with a single CPU thread
✅SOTA in monocular 3D reconstruction

More: https://bit.ly/37pQS40

🪰NUWA-Infinity is out!🪰

👉∞ generation by #Microsoft: arbitrarily-sized HD images and long videos 🤯

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅Unconditional Image Gen.
✅Text-to-Image/Text-to-Clip
✅Animation / Out-painting
✅Hi-res, arbitrary long clip
✅NCP for patches caching

More: https://bit.ly/3zmBf9f

🧰 FGT: flow-guided inpainting 🧰

👉#Microsoft (+USTC) unveils FGT: flow-guided ViT for video inpainting 🤯

𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬:
✅OF into transformer for attention++
✅Flow completion net w/ local feats.
✅Dual perspective spatial MHSA
✅Local attention with global content

More: https://bit.ly/3pk5J5S

😈 Synthetic Expression-Wrinkles 😈

👉#Microsoft unveils a novel approach that produces realistic wrinkles across humans

😎Review https://bit.ly/3zWZLOd
😎Paper arxiv.org/pdf/2210.03529.pdf
😎Project microsoft.github.io/DynamicWrinkles

🪴 Rodin: 3D Avatars Using Diffusion 🪴

👉#Microsoft unveils a novel #3D diffusion for digital avatars as NeRF

😎Review https://bit.ly/3jcxeOX
😎Project 3d-avatar-diffusion.microsoft.com
😎Paper arxiv.org/pdf/2212.06135.pdf

🗣️ MemFace: Generative Talking Face 🗣️

👉#Microsoft (+SJTU) unveils MemFace: the new SOTA in talking faces generation

😎Review https://bit.ly/3k8TjhZ
😎Paper arxiv.org/pdf/2212.05005v2.pdf
😎Project memoryface.github.io/

🪩 DISCO: Human Dance Generation 🪩

👉NTU (+ #Microsoft) unveils DISCO: a big step towards the Human Dance Generation.

😎Review https://t.ly/cNGX
😎Paper arxiv.org/pdf/2307.00040.pdf
😎Project disco-dance.github.io/
😎Code github.com/Wangt-CN/DisCo

🍉 AltFreezing: new SOTA in detecting deepfake 🍉

👉#Microsoft unveils AltFreezing: spatial/temporal artifacts in one model for more general face forgery detection

😎Review https://t.ly/mkIKX
😎Paper https://t.ly/z4KnJ
😎Code github.com/ZhendongWang6/AltFreezing

🍄 Video Understanding with GPT-4V(ision) 🍄

👉 #Microsoft unveils MM-Vid, the most advanced video understanding framework (w/ #chatgpt4). Impressive results on long-form videos & intricate tasks such as audio description & multimodal high-level comprehension

😎Review https://t.ly/RISMm
😎Paper arxiv.org/pdf/2310.19773.pdf
😎Project https://multimodal-vid.github.io

🔥Florence-2: unified Computer Vision🔥

👉#Microsoft announces Florence-2: novel foundation model with unified, prompt-based, representation for a large variety of #computervision & vision-language task. One backbone -> multiple tasks!

👉Review https://t.ly/pOins
👉Paper arxiv.org/pdf/2311.06242.pdf
👉Project www.microsoft.com/en-us/research/project/projectflorence/