zju3dv / neuralbody

Code for "Neural Body: Implicit Neural Representations with Structured Latent Codes for Novel View Synthesis of Dynamic Humans" CVPR 2021 best paper candidate
Other
907 stars 130 forks source link

请问如何将多个RGB-D cameras捕获的自定义pose信息运用到Neuralbody上 #144

Closed felixshing closed 1 year ago

felixshing commented 1 year ago

你好,我们这边在做一个live interactive volumetric video系统。每一边用户会有3-4个Zed 2i depth cameras捕获动作。我们现在的想法是在头几个frame用neuralbody把human的NeRF representation表示出来。然后接下来的frame我们只传pose的变化。

为了让重建的人物更逼真,我在考虑从人体身上捕获更多的keypoints,比如zed sdk提供70个keypoints: https://www.stereolabs.com/docs/body-tracking/; Google的MediaPipe可以在脸上提供487个keypoints: https://developers.google.com/mediapipe/solutions/vision/face_landmarker

请问如果我希望用这么多keypoints的话,我怎么样调用Neuralbody呢