An interesting book (2013). Some may think there are BS points in it, but for some of you it might be an interesting read.
Excerpts:
"To consider the problem of generating motions for a character in a movie, it is important to realize that the motions can be created procedurally, i.e. by designing algorithms that synthesize motion. The animations can be created "by hand" or captured from an actor in a studio. These "pure data" approaches give the highest quality motions, but at substantial cost in time and effort of artists or actors. Moreover, there is little flexibility: If it is discovered that the right motions were not captured in the studio, it is necessary to retrack and capture more. The situation is worse for a video game, where all the motions that might conceivably be needed must be captured. To solve this problem, machine learning techniques can be adopted to promise the best of both worlds: Starting with an amount of captured data, we can procedurally synthesize more data in the style of the original. Moreover, we can constrain the synthetic data, for example, according to the requirements of an artist. For such problems, machine learning offers an attractive set of tools for modeling the patterns of data. These data-driven techniques have gained a steadily increasing presence in graphics research."
"Computer animation is a time-intensive process, and this is true for both 2D and 3D cartoon animation. As a result, there has long been interest in partially automating animation, and indeed, animation has thus far benefited greatly from applications of machine learning. The majority of the computer animation production burden is usually artistic, not scientific or computational. This is due to the fact that computer animation algorithms are eminently reusable, but the data they operate on is often custom-designed and highly stylized. This bottleneck is illustrated by computer-generated movie production in which the human workload is usually more than 80 % artistic (e.g., modeling, texturing, animating, etc.).
There is therefore a need in computer animation for data transformation and modeling techniques that can synthesize and/or generalize data, thereby at least partially alleviating the data bottleneck. Machine learning techniques are proposed to fulfill this need, and they have two functions that are particularly useful to computer graphics: (1) They extract functional information from data, and (2) they synthesize new data based on existing data. Machine learning allows us to leverage existing data in a nondirect and nontrivial manner, which can save both human and computational time. For example, techniques have been developed for generating meshes of novel human bodies given a small set of example meshes. This is done by creating a generative model of the mesh data through regression."
Description
The integration of machine learning techniques and cartoon animation research is fast becoming a hot topic. This book helps readers learn the latest machine learning techniques, including patch alignment framework; spectral clustering, graph cuts, and convex relaxation; ensemble manifold learning; multiple kernel learning; multiview subspace learning; and multiview distance metric learning. It then presents the applications of these modern machine learning techniques in cartoon animation research. With these techniques, users can efficiently utilize the cartoon materials to generate animations in areas such as virtual reality, video games, animation films, and sport simulations.
Contents
1 Introduction
1.1 Perception
1.2 Overview of Machine Learning Techniques
1.2.1 Manifold Learning
1.2.2 Semi-supervised Learning
1.2.3 Multiview Learning
1.2.4 Learning-based Optimization
1.3 Recent Developments in Computer Animation
1.3.1 Example-Based Motion Reuse
1.3.2 Physically Based Computer Animation
1.3.3 Computer-Assisted Cartoon Animation
1.3.4 Crowd Animation
1.3.5 Facial Animation
2 Modern Machine Learning Techniques
2.1 A Unified Framework for Manifold Learning
2.1.1 Framework Introduction
2.1.2 Various Manifold Learning Algorithm Unifying
2.1.3 Discriminative Locality Alignment
2.2 Spectral Clustering and Graph Cut
2.2.1 Spectral Clustering
2.2.2 Graph Cut Approximation
2.3 Ensemble Manifold Learning
2.3.1 Motivation for EMR
2.3.2 Overview of EMR
2.3.3 Applications of EMR
2.4 Multiple Kernel Learning
2.4.1 A Unified Mulitple Kernel Learning Framework
2.4.2 SVM with Multiple Unweighted-Sum Kernels
2.4.3 QCQP Multiple Kernel Learning
2.5 Multiview Subspace Learning
2.5.1 Approach Overview
2.5.2 Techinique Details
2.5.3 Alternative Optimization Used in PA-MSL
2.6 Multiview Distance Metric Learning
2.6.1 Motivation for MDML
2.6.2 Graph-Based Semi-supervised Learning
2.6.3 Overview of MDML
2.7 Multi-task Learning
2.7.1 Introduction of Structural Learning
2.7.2 Hypothesis Space Selection
2.7.3 Algorithm for Multi-task Learning
2.7.4 Solution by Alternative Optimization
3 Animation Research: A Brief Introduction
3.1 Traditional Animation Production
3.1.1 History of Traditional Animation Production
3.1.2 Procedures of Animation Production
3.1.3 Relationship Between Traditional Animation and Computer Animation
3.2 Computer-Assisted Systems
3.2.1 Computer Animation Techniques
3.3 Cartoon Reuse Systems for Animation Synthesis
3.3.1 Cartoon Texture for Animation Synthesis
3.3.2 Cartoon Motion Reuse
3.3.3 Motion Capture Data Reuse in Cartoon Characters
3.4 Graphical Materials Reuse: More Examples
3.4.1 Video Clips Reuse
3.4.2 Motion Captured Data Reuse by Motion Texture
3.4.3 Motion Capture Data Reuse by Motion Graph
4 Animation Research: Modern Techniques
4.1 Automatic Cartoon Generation with Correspondence Construction
4.1.1 Related Work in Correspondence Construction
4.1.2 Introduction of the Semi-supervised Correspondence Construction
4.1.3 Stroke Correspondence Construction via Stroke Reconstruction Algorithm
4.1.4 Simulation Results
4.2 Cartoon Characters Represented by Multiple Features
4.2.1 Cartoon Character Extraction
4.2.2 Color Histogram
4.2.3 Hausdorff Edge Feature
4.2.4 Motion Feature
4.2.5 Skeleton Feature
4.2.6 Complementary Characteristics of Multiview Features
4.3 Graph-based Cartoon Clips Synthesis
4.3.1 Graph Model Construction
4.3.2 Distance Calculation
4.3.3 Simulation Results
4.4 Retrieval-based Cartoon Clips Synthesis
4.4.1 Constrained Spreading Activation Network
4.4.2 Semi-supervised Multiview Subspace Learning
4.4.3 Simulation Results
https://www.amazon.com/Learning-Techniques-Applications-Animation-Research/dp/1118115147
zov |||