Excited to share our #CVPR2020 paper on "Three-Dimensional Reconstruction of Human Interactions". We model and learn physical contact between humans for more accurate 3d reconstruction of interacting people. Project Page: vision.imar.ro/ci3d PDF: openaccess.thecvf.com/content_CVPR_2…

#MediaPipe BlazePose GHUM 3D (goo.gle/2PObiaO) is now 3x more accurate, achieving state-of-the-art performance while offering new lite and heavy versions to cover every device and use case! Try it out at solutions.mediapipe.dev/pose for Python, JS, Android and iOS!

Our statistical 3d human pose and shape models (GHUM) are available at github.com/google-researc….

BlazePoseGHUM3D is now available (Blog Post developers.googleblog.com/2021/03/ml-kit…) in MLKit and MediaPipe (google.github.io/mediapipe/solu…)

New monocular (full-body) 3d human pose and shape estimation (a.k.a. HUND) Neural Descent for Visual 3D Human Pose and Shape Andrei Zanfir, Eduard Gabriel Bazavan, Mihai Zanfir, William T. Freeman, Rahul Sukthankar, Cristian Sminchisescu, CVPR2021.

Join us in 10 minutes (10:55 am EDT, 16:55 cet) at #CVPR2021 and exercise! Find out about our AIFIT: Automatic 3D Human-Interpretable Feedback Models for Fitness Training. Paper/data: vision.imar.ro/fit3d/.