What this book covers
Chapter 1, Introducing 3D Data Processing, will cover the basics of 3D data, such as how 3D data is stored and the basic concepts of meshes and point clouds, world coordinations, and camera coordinations. It also shows us what NDC is (a frequently used coordination), how to convert between different coordinations, perspective cameras, and orthographic cameras, and which camera models should be used.
Chapter 2, Introducing 3D Computer Vision and Geometry, will show us the basic concepts in computer graphics, such as rendering and shading. We will learn about some fundamental concepts that will be required in the later chapters of this book, including, 3D geometry transforms, PyTorch tensors, and optimization.
Chapter 3, Fitting Deformable Mesh Models to Raw Point Clouds, will present a hands-on project of using a deformable 3D model to fit a noisy 3D observation using all the knowledge that we have learned in the previous chapters. We will explore frequently used cost functions, why these cost functions are important, and when these cost functions are usually used. Finally, we will explore a concrete example of which cost functions have been selected for which tasks and how to set up the optimization loop to obtain the results that we want.
Chapter 4, Learning Object Pose Detection and Tracking by Differentiable Rendering, will talk about the basic concepts of differentiable rendering. It will help you understand the basic concepts and know when you can apply these techniques to solve your own problems.
Chapter 5, Understanding Differentiable Volumetric Rendering, will present a hands-on project using differentiable rendering to estimate camera positions from a single image and a known 3D mesh model. We will learn how to practically use PyTorch3D to set up cameras, renders, and shaders. We will also get hands-on experience in using different cost functions to get optimization results.
Chapter 6, Exploring Neural Radiance Fields (NeRF), will provide a hands-on project using differentiable rendering to estimate 3D mesh models from several images and texture models.
Chapter 7, Exploring Controllable Neural Feature Fields, will cover a very important algorithm for view synthesis, which is NeRF. We will learn what it is all about, how to use it, and where it is valuable.
Chapter 8, Modeling the Human Body in 3D, will explore 3D human body fitting using the SMPL algorithm.
Chapter 9, Performing End-to-End View Synthesis with SynSin, will cover SynSin, which is a state-of-the-art deep learning image synthesis model.
Chapter 10, Mesh R-CNN, will introduce us to Mesh R-CNN, which is another state-of-the-art method for predicting 3D voxel models from a single input image.
