Computer Vision -- ECCV 2020 : 16th European Conference, Glasgow, UK, August 23-28, 2020, Proceedings. Part V / Andrea Vedaldi, Horst Bischof, Thomas Brox, Jan-Michael Frahm (eds.).

Intro
Foreword
Preface
Organization
Contents
Part V
BLSM: A Bone-Level Skinned Model of the Human Mesh
1 Introduction
2 Related Work
3 Bone-Level Skinned Model
3.1 Skeleton Modeling
3.2 Template Synthesis
3.3 Linear Blend Skinning
4 Model Training
4.1 Unconstrained Landmark-Based Alignment
4.2 Bone Basis and Bone-Corrective Blendshapes
4.3 Shape Blendshapes
4.4 Blending Weights
5 Evaluation
5.1 Implementation Details
5.2 Quantitative Evaluation
5.3 Qualitative Evaluation
6 Conclusion
References

Associative Alignment for Few-Shot Image Classification
1 Introduction
2 Related Work
3 Preliminaries
4 Associative Alignment
4.1 Detecting the Related Bases
4.2 Centroid Associative Alignment
4.3 Adversarial Associative Alignment
5 Establishing a Strong Baseline
5.1 Classification Loss Functions
5.2 Episodic Early Stopping
6 Experimental Validation
6.1 Datasets and Implementation Details
6.2 mini-ImageNet and CUB with a Shallow Conv4 Backbone
6.3 mini-ImageNet and tieredimageNet with Deep Backbones
6.4 FC100 and CUB with a ResNet-18 Backbone

6.5 Cross-Domain Evaluation
7 Discussion
References
Cyclic Functional Mapping: Self-supervised Correspondence Between Non-isometric Deformable Shapes
1 Introduction
2 Contribution
3 Background
3.1 Riemannian 2-Manifolds
3.2 Functional Maps
3.3 Deep Functional Maps
3.4 Self-Supervised Deep Functional Maps
4 Cyclic Self-Supervised Deep Functional Maps
4.1 Correspondence Distortion
4.2 Cyclic Distortion
4.3 Deep Cyclic Mapping
5 Implementation
5.1 Hardware
5.2 Pre-processing
5.3 Network Architecture
6 Experiments
6.1 Mesh Error Evaluation

6.2 Synthetic FAUST
6.3 Real Scans
6.4 Non-Isometric Deformations
6.5 TOSCA
6.6 SCAPE
6.7 One-Shot Single Pair Learning
6.8 Partial Shapes Correspondence
7 Limitations
8 Summary
References
View-Invariant Probabilistic Embedding for Human Pose
1 Introduction
2 Related Work
3 Our Approach
3.1 Matching Definition
3.2 Triplet Ratio Loss
3.3 Positive Pairwise Loss
3.4 Probabilistic Embeddings
3.5 Camera Augmentation
3.6 Implementation Details
4 Experiments
4.1 Datasets
4.2 View-Invariant Pose Retrieval
4.3 Downstream Tasks

4.4 Ablation Study
5 Conclusion
References
Contact and Human Dynamics from Monocular Video
1 Introduction
2 Related Work
3 Physics-Based Motion Estimation
3.1 Physics-Based Trajectory Optimization
3.2 Learning to Estimate Contacts
3.3 Kinematic Initialization
4 Results
4.1 Contact Estimation
4.2 Qualitative Motion Evaluation
4.3 Quantitative Motion Evaluation
5 Discussion
References
PointPWC-Net: Cost Volume on Point Clouds for (Self-)Supervised Scene Flow Estimation
1 Introduction
2 Related Work
3 Approach
3.1 The Cost Volume Layer