As this ebook edition doesn't have fixed pagination, the page numbers below are hyperlinked for reference only, based on the printed edition of this book.
Symbols
3D data file format
3D data representation 5
mesh representation 6
point cloud representation 5
3D human pose and shape
estimating, with SMPLify 160
3D modeling problem
formulating 152
representation, defining 153, 154
3D models
reconstructing, from multi-view images 104-109
3D rendering
A
Anaconda
URL 4
B
barycentric coordinates 31, 32
blend shapes 155
C
Chamfer distance 57
CMU Graphics Lab Motion Capture Database 161
compositional 3D-aware image synthesis 135-138
compositional operator 136
controllable car generation
controllable face generation
controllable scene generation
exploring 141
Convolutional Neural Network (CNN) 136
coordination systems
cost function 57
D
deformable mesh fitting problem
formulating, into optimization problem 56, 57
development environment
differentiable rendering
need for 68-70
used, for solving issues 72
differentiable volume rendering 104
3D models, reconstructing from multi-view images 104-109
E
exponential map 48
F
feature field 139
generating 138
Frechet Inception Distance 147
G
GAN-based image synthesis 134, 135
GAN model architecture 136
Generative Adversarial Networks (GANs) 134
GIRAFFE model 135
compositional operator 136
Frechet Inception Distance 147
GAN model architecture 136
learning 3D representation 136
neural rendering model 136
pre-trained dataset models 141
GIRAFFE model, training parameters
data 148
model 148
training 148
graph convolutional networks (GCNs) 194
H
hat operator 48
heterogeneous mini-batch
list format 44
packed format 44
padded format 44
I
identity-based shape 156
images
feature field, mapping to 139, 141
L
Lambertian shading model 33
Laplacian 58
learning 3D representation 136
light source models 32
Linear Blend Skinning technique 154, 155
logarithmic map 48
loss functions 55
properties requisites 55
M
Material Template Library (MTL) file 13
mesh
mesh edge loss 59
usage, experimenting 64
meshes, fitting to point clouds
problem 55
mesh fitting
implementing, with PyTorch3D 59-63
mesh Laplacian smoothing loss 58
mesh normal consistency loss 58
Mesh R-CNN
Mesh R-CNN architecture 191-194
mesh representation 6
Multi-Layer Perceptron (MLP) 114
multi-view images
3D models, reconstructing from 104-109
N
NeRF model 114
neural networks
used, for representing radiance fields 113
neural point cloud renderer
versus naïve renderers 173
Neural Radiance Fields (NeRF) 111-133
neural rendering model 136
Normalized Device Coordinates (NDCs) 117
O
object pose estimation
example, for silhouette fitting 84-93
example, for texture fitting 84-93
optimization objective function
optimization problem
deformable mesh fitting problem, formulating into 56, 57
P
point cloud representation 5
PointNet
URL 5
PyTorch
Mesh R-CNN, using with 198-206
PyTorch3D
used, for implementing mesh fitting 59-63
PyTorch3D heterogeneous batches
PyTorch optimizers
R
representing, with neural networks 113
volume rendering 128
ray marcher
ray marching 97
rays
projecting, into scene 129
region proposal network (RPN) 194
regularization 58
regularization loss functions 57
mesh edge loss 59
mesh Laplacian smoothing loss 58
mesh normal consistency loss 58
mesh, obtained with mesh edge loss 64
mesh, without regularization loss functions 63
representation
S
scene
rays, projecting into 129
shape 156
shape-based shape 156
Signed Distance Function (SDF) 6
silhouette fitting
object pose estimation, example 84-93
Skinned Multi-Person Linear (SMPL) model 156
defining 156
pose-dependent template mesh 157
shape dependent joints 157
skinning 154
SMPLify
exploring 161
stages 160
used, for estimating 3D human pose and shape 160
stereo vision cameras 54
Stochastic Gradient Descent (SGD) 42
Structural Similarity Index (SSIM) 179
SynSin model
network architecture 171
SynSin network architecture 171
depth network 172
discriminator 176
neural point cloud renderer 172-175
spatial feature 172
T
texture fitting
object pose estimation, example 84-93
time-of-flight cameras 54
Truncated Signed Distance Functions (TSDFs) 6
V
view synthesis 112
challenges 170
methods 171
volume 96
volume (volumetric) rendering 128
with radiance fields 128
W
world coordination system 22
3.135.198.49