microsoft / Deep3DFaceReconstruction

Accurate 3D Face Reconstruction with Weakly-Supervised Learning: From Single Image to Image Set (CVPRW 2019)
MIT License
2.16k stars 441 forks source link

pixel normal #212

Open chen990627 opened 1 year ago

chen990627 commented 1 year ago

`pixel_attributes,alphas = rasterize_triangles.rasterize_triangles( vertices, vertex_attributes, triangles, clip_space_transforms, image_width, image_height, [-1] * vertex_attributes.shape[2])

pixel_normals = tf.nn.l2_normalize(pixel_attributes[:, :, :, 0:3], axis=3)`

the code above is from mesh_renderer.py and I debug it with a image. why does the pixel normal contain negative values? what do these negative values represent?