Channel name was changed to ยซDeep Learningยป
Channel name was changed to ยซAI & Deep Learningยป
๐ฃ๐ฅ#Unity acquires Ziva Dynamics ๐ฃ๐ฅ
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ Real-time digital humans
โ A mix of #ML & biomechanics
โ Complex anatomical simulation
โ 30 terabytes of 4D data
โ 72,000 trained shapes
More: https://www.linkedin.com/posts/visionarynet_unity-metaverse-ml-activity-6891413170834030592-Rr9p
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ Real-time digital humans
โ A mix of #ML & biomechanics
โ Complex anatomical simulation
โ 30 terabytes of 4D data
โ 72,000 trained shapes
More: https://www.linkedin.com/posts/visionarynet_unity-metaverse-ml-activity-6891413170834030592-Rr9p
Linkedin
#unity #metaverse #ml #deeplearning #ai #ai #artificialintelligenceโฆ | Alessandro Ferrari | 22 comments
๐ฃ๐ฅ#Unity acquires Ziva Dynamics ๐ฃ๐ฅ
๐Unity acquires Ziva Dynamics, leader in sophisticated simulation and deformation, machine learning, and real-time character creation. #Metaverse?
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ Real-time SOTA in Unity for digital humans creation
โ A mixโฆ
๐Unity acquires Ziva Dynamics, leader in sophisticated simulation and deformation, machine learning, and real-time character creation. #Metaverse?
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ Real-time SOTA in Unity for digital humans creation
โ A mixโฆ
๐ค1๐ฑ1
๐ง๐งThe definitive (recently updated) collection of neural pedestrian detectors๐ง๐ง
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ Cascade Mask-R-CNN
โ Faster R-CNN
โ RetinaNet(Guided Anchoring)
โ Hybrid Task Cascade, MGAN, CSP
โ MMDetection/SwinTransformer
โ Code under Apache License 2.0
More: https://bit.ly/3KArsA1
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ Cascade Mask-R-CNN
โ Faster R-CNN
โ RetinaNet(Guided Anchoring)
โ Hybrid Task Cascade, MGAN, CSP
โ MMDetection/SwinTransformer
โ Code under Apache License 2.0
More: https://bit.ly/3KArsA1
๐ฅ2๐คฉ2๐1๐ฑ1
๐ง๐งTraining ImageNet in minutes with FFCV๐ง๐ง
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ Work with any existing code
โ Pre-fetching, caching, transfer
โ Fused-and-compiled data proc.
โ Multiple models per GPU
โ Tools for image handling
More: https://bit.ly/3IAZHW9
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ Work with any existing code
โ Pre-fetching, caching, transfer
โ Fused-and-compiled data proc.
โ Multiple models per GPU
โ Tools for image handling
More: https://bit.ly/3IAZHW9
๐3๐ฅ1๐คฉ1
This media is not supported in your browser
VIEW IN TELEGRAM
๐จRePaint: new SOTA in inpainting๐จ
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ RePaint via DDPM
โ Suitable with severe corruption
โ Pretrained DDPM as prior
โ SOTA vs Autoregressive/GAN
More: https://bit.ly/3AAk8jm
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ RePaint via DDPM
โ Suitable with severe corruption
โ Pretrained DDPM as prior
โ SOTA vs Autoregressive/GAN
More: https://bit.ly/3AAk8jm
๐คฉ4๐คฏ1
Channel name was changed to ยซAI & Deep Learning (with papers)ยป
This media is not supported in your browser
VIEW IN TELEGRAM
๐ ๐ Re-inventing the animation๐ ๐
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ S2PR: engineers + artists
โ Style2Paint: a coloring software
โ Style2Paints -> โSePaโ
โ Source code under Apache
More: https://bit.ly/3rPZPu4
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ S2PR: engineers + artists
โ Style2Paint: a coloring software
โ Style2Paints -> โSePaโ
โ Source code under Apache
More: https://bit.ly/3rPZPu4
๐2
This media is not supported in your browser
VIEW IN TELEGRAM
๐ฝ ManyDepth: adaptive 3D-depth ๐ฝ
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ SSL-training, monocular only
โ No depths or poses needed
โ Adaptive cost volume
โ Handling moving objects
โ Patent Pending๐
More: https://bit.ly/3fZpIlD
๐๐ข๐ ๐ก๐ฅ๐ข๐ ๐ก๐ญ๐ฌ:
โ SSL-training, monocular only
โ No depths or poses needed
โ Adaptive cost volume
โ Handling moving objects
โ Patent Pending๐
More: https://bit.ly/3fZpIlD
๐4โค2๐ฅ1๐ฑ1