Real-time 3D face mesh from one camera on phones
Real-time 3D face mesh from one camera on phones
Real-time Facial Surface Geometry from Monocular Video on Mobile GPUs
arXiv paper abstract https://arxiv.org/abs/1907.06724
arXiv PDF paper https://arxiv.org/pdf/1907.06724.pdf
PyTorch GitHub https://github.com/thepowerfuldeez/facemesh.pytorch
TensorFlow GitHub https://github.com/gouthamvgk/facemesh_coreml_tf
Project Web page https://sites.google.com/view/perception-cv4arvr/facemesh
Project demo https://gradio.app/g/AK391/FaceMesh
… inferring an approximate 3D mesh representation of a human face from single camera input for AR applications.
… dense mesh model of 468 vertices is well-suited for face-based AR effects.
… inference speed on mobile GPUs (100–1000+ FPS …)
… prediction quality that is comparable to the variance in manual annotations of the same image.
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
