πHyper-Dense Landmarks at 150FPSπ
π#Microsoft unveils the SOTA in dense landmarking + #3D reconstruction. MAGIC.
ππ’π π‘π₯π’π π‘ππ¬:
β Accurate 10Γ as many landmarks as usual
β Synthetic data, perfect annotations
β NO appearance, light, diff-rendering
β #3D @150+FPS with a single CPU thread
β SOTA in monocular 3D reconstruction
More: https://bit.ly/37pQS40
π#Microsoft unveils the SOTA in dense landmarking + #3D reconstruction. MAGIC.
ππ’π π‘π₯π’π π‘ππ¬:
β Accurate 10Γ as many landmarks as usual
β Synthetic data, perfect annotations
β NO appearance, light, diff-rendering
β #3D @150+FPS with a single CPU thread
β SOTA in monocular 3D reconstruction
More: https://bit.ly/37pQS40
π6π₯4π€―1
This media is not supported in your browser
VIEW IN TELEGRAM
πͺ°NUWA-Infinity is out!πͺ°
πβ generation by #Microsoft: arbitrarily-sized HD images and long videos π€―
ππ’π π‘π₯π’π π‘ππ¬:
β Unconditional Image Gen.
β Text-to-Image/Text-to-Clip
β Animation / Out-painting
β Hi-res, arbitrary long clip
β NCP for patches caching
More: https://bit.ly/3zmBf9f
πβ generation by #Microsoft: arbitrarily-sized HD images and long videos π€―
ππ’π π‘π₯π’π π‘ππ¬:
β Unconditional Image Gen.
β Text-to-Image/Text-to-Clip
β Animation / Out-painting
β Hi-res, arbitrary long clip
β NCP for patches caching
More: https://bit.ly/3zmBf9f
π₯7π2β€1π1π€―1
This media is not supported in your browser
VIEW IN TELEGRAM
π§° FGT: flow-guided inpainting π§°
π#Microsoft (+USTC) unveils FGT: flow-guided ViT for video inpainting π€―
ππ’π π‘π₯π’π π‘ππ¬:
β OF into transformer for attention++
β Flow completion net w/ local feats.
β Dual perspective spatial MHSA
β Local attention with global content
More: https://bit.ly/3pk5J5S
π#Microsoft (+USTC) unveils FGT: flow-guided ViT for video inpainting π€―
ππ’π π‘π₯π’π π‘ππ¬:
β OF into transformer for attention++
β Flow completion net w/ local feats.
β Dual perspective spatial MHSA
β Local attention with global content
More: https://bit.ly/3pk5J5S
β€11π5
This media is not supported in your browser
VIEW IN TELEGRAM
π Synthetic Expression-Wrinkles π
π#Microsoft unveils a novel approach that produces realistic wrinkles across humans
πReview https://bit.ly/3zWZLOd
πPaper arxiv.org/pdf/2210.03529.pdf
πProject microsoft.github.io/DynamicWrinkles
π#Microsoft unveils a novel approach that produces realistic wrinkles across humans
πReview https://bit.ly/3zWZLOd
πPaper arxiv.org/pdf/2210.03529.pdf
πProject microsoft.github.io/DynamicWrinkles
π₯7π€―4π2π±1
This media is not supported in your browser
VIEW IN TELEGRAM
πͺ΄ Rodin: 3D Avatars Using Diffusion πͺ΄
π#Microsoft unveils a novel #3D diffusion for digital avatars as NeRF
πReview https://bit.ly/3jcxeOX
πProject 3d-avatar-diffusion.microsoft.com
πPaper arxiv.org/pdf/2212.06135.pdf
π#Microsoft unveils a novel #3D diffusion for digital avatars as NeRF
πReview https://bit.ly/3jcxeOX
πProject 3d-avatar-diffusion.microsoft.com
πPaper arxiv.org/pdf/2212.06135.pdf
β€9π€―4π2π±1
This media is not supported in your browser
VIEW IN TELEGRAM
π£οΈ MemFace: Generative Talking Face π£οΈ
π#Microsoft (+SJTU) unveils MemFace: the new SOTA in talking faces generation
πReview https://bit.ly/3k8TjhZ
πPaper arxiv.org/pdf/2212.05005v2.pdf
πProject memoryface.github.io/
π#Microsoft (+SJTU) unveils MemFace: the new SOTA in talking faces generation
πReview https://bit.ly/3k8TjhZ
πPaper arxiv.org/pdf/2212.05005v2.pdf
πProject memoryface.github.io/
π€―12π€©3π1
This media is not supported in your browser
VIEW IN TELEGRAM
πͺ© DISCO: Human Dance Generation πͺ©
πNTU (+ #Microsoft) unveils DISCO: a big step towards the Human Dance Generation.
πReview https://t.ly/cNGX
πPaper arxiv.org/pdf/2307.00040.pdf
πProject disco-dance.github.io/
πCode github.com/Wangt-CN/DisCo
πNTU (+ #Microsoft) unveils DISCO: a big step towards the Human Dance Generation.
πReview https://t.ly/cNGX
πPaper arxiv.org/pdf/2307.00040.pdf
πProject disco-dance.github.io/
πCode github.com/Wangt-CN/DisCo
π₯13π₯°4π2β‘1π1πΎ1
This media is not supported in your browser
VIEW IN TELEGRAM
π AltFreezing: new SOTA in detecting deepfake π
π#Microsoft unveils AltFreezing: spatial/temporal artifacts in one model for more general face forgery detection
πReview https://t.ly/mkIKX
πPaper https://t.ly/z4KnJ
πCode github.com/ZhendongWang6/AltFreezing
π#Microsoft unveils AltFreezing: spatial/temporal artifacts in one model for more general face forgery detection
πReview https://t.ly/mkIKX
πPaper https://t.ly/z4KnJ
πCode github.com/ZhendongWang6/AltFreezing
π±6π5π4π€―2π₯°1
This media is not supported in your browser
VIEW IN TELEGRAM
π Video Understanding with GPT-4V(ision) π
π #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
πReview https://t.ly/RISMm
πPaper arxiv.org/pdf/2310.19773.pdf
πProject https://multimodal-vid.github.io
π #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
πReview https://t.ly/RISMm
πPaper arxiv.org/pdf/2310.19773.pdf
πProject https://multimodal-vid.github.io
π€―22π9π₯2π1π±1
This media is not supported in your browser
VIEW IN TELEGRAM
π₯Florence-2: unified Computer Visionπ₯
π#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!
πReview https://t.ly/pOins
πPaper arxiv.org/pdf/2311.06242.pdf
πProject www.microsoft.com/en-us/research/project/projectflorence/
π#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!
πReview https://t.ly/pOins
πPaper arxiv.org/pdf/2311.06242.pdf
πProject www.microsoft.com/en-us/research/project/projectflorence/
π±9β€5π₯3π1π1πΎ1
This media is not supported in your browser
VIEW IN TELEGRAM
π©° Dressed Humans in the wild π©°
πETH (+ #Microsoft ) ReLoo: novel 3D-HQ reconstruction of humans dressed in loose garments from mono in-the-wild clips. No prior assumptions about the garments. Source Code announced, coming π
πReview https://t.ly/evgmN
πPaper arxiv.org/pdf/2409.15269
πProject moygcc.github.io/ReLoo/
πCode github.com/eth-ait/ReLoo
πETH (+ #Microsoft ) ReLoo: novel 3D-HQ reconstruction of humans dressed in loose garments from mono in-the-wild clips. No prior assumptions about the garments. Source Code announced, coming π
πReview https://t.ly/evgmN
πPaper arxiv.org/pdf/2409.15269
πProject moygcc.github.io/ReLoo/
πCode github.com/eth-ait/ReLoo
π€―9β€2π1π₯1
This media is not supported in your browser
VIEW IN TELEGRAM
π₯BitNet: code of 1-bit LLM releasedπ₯
πBitNet by #Microsoft, announced in late 2023, is a 1-bit Transformer architecture designed for LLMs. BitLinear as a drop-in replacement of the nn.Linear layer in order to train 1-bit weights from scratch. Source Code just released π
πReview https://t.ly/3G2LA
πPaper arxiv.org/pdf/2310.11453
πCode https://lnkd.in/duPADJVb
πBitNet by #Microsoft, announced in late 2023, is a 1-bit Transformer architecture designed for LLMs. BitLinear as a drop-in replacement of the nn.Linear layer in order to train 1-bit weights from scratch. Source Code just released π
πReview https://t.ly/3G2LA
πPaper arxiv.org/pdf/2310.11453
πCode https://lnkd.in/duPADJVb
π₯21β€5π€―2π1π₯°1
This media is not supported in your browser
VIEW IN TELEGRAM
π§Ώ Look Ma, no markers π§Ώ
π#Microsoft unveils the first technique for marker-free, HQ reconstruction of COMPLETE human body, including eyes and tongue, without requiring any calibration, manual intervention or custom hardware. Impressive results! Repo for training & Dataset releasedπ
πReview https://t.ly/5fN0g
πPaper arxiv.org/pdf/2410.11520
πProject microsoft.github.io/SynthMoCap/
πRepo github.com/microsoft/SynthMoCap
π#Microsoft unveils the first technique for marker-free, HQ reconstruction of COMPLETE human body, including eyes and tongue, without requiring any calibration, manual intervention or custom hardware. Impressive results! Repo for training & Dataset releasedπ
πReview https://t.ly/5fN0g
πPaper arxiv.org/pdf/2410.11520
πProject microsoft.github.io/SynthMoCap/
πRepo github.com/microsoft/SynthMoCap
π€―16π10π₯3π±3β€1π1
This media is not supported in your browser
VIEW IN TELEGRAM
πDAViD: Synthetic Depth-Normal-Segmentationπ
π#Microsoft's DAViD: 100% synthetic dataset/models for human Depth, Normals & Segmentation. Dataset available, models & runtime under MITπ
πReview https://t.ly/-SlO_
πPaper https://lnkd.in/eCmMXpTg
πProject https://lnkd.in/eurCSWkm
πRepo https://lnkd.in/e7PWFgP2
π#Microsoft's DAViD: 100% synthetic dataset/models for human Depth, Normals & Segmentation. Dataset available, models & runtime under MITπ
πReview https://t.ly/-SlO_
πPaper https://lnkd.in/eCmMXpTg
πProject https://lnkd.in/eurCSWkm
πRepo https://lnkd.in/e7PWFgP2
π6β€4π₯2π€©1