Intro
Preface
Organizing Committee
Contents - Part I
Contents - Part II
Detection, Recognition and Identification
MMM-GCN: Multi-Level Multi-Modal Graph Convolution Network for Video-Based Person Identification
1 Introduction
2 Related Work
2.1 Video-Based Multi-Modal Person Identification
2.2 Graph Convolution Networks
3 Methodology
3.1 Overview
3.2 Graph Construction and Learning
3.3 Graph Node Encoding
3.4 Person Identification
4 Experiment
4.1 Experiment Setups
4.2 Comparison to the State of the Art
4.3 Ablation Study

4.4 Discussion
5 Conclusion
References
Feature Enhancement and Reconstruction for Small Object Detection
1 Introduction
2 Related Work
2.1 Upsampling Method
2.2 Multi-scale Feature Extraction
2.3 Attention Mechanism
3 Method
3.1 Network Architecture
3.2 Content-Aware Upsampling (CAU)
3.3 Channel Shuffle Attention (CSA)
4 Experiments
4.1 Datasets
4.2 Evaluation Metrics
4.3 Implementation Details
4.4 Ablation Study
4.5 Comparative Results
4.6 Qualitative Results
5 Conclusion
References

Toward More Accurate Heterogeneous Iris Recognition with Transformers and Capsules
1 Introduction
2 Technical Details
2.1 Vision Token Generator
2.2 Transformer Encoder
2.3 Transformer Decoder
2.4 3D Capsule Matcher
2.5 Loss Function
3 Experiments and Results
3.1 Datasets
3.2 Network Structure
3.3 Training Setup
3.4 Evaluation Protocol
3.5 Comparison with Other State-of-the-art Heterogeneous Iris Recognition Algorithms
3.6 Ablation Experiments
4 Conclusions
References

MCANet: Multiscale Cross-Modality Attention Network for Multispectral Pedestrian Detection
1 Introduction
2 Related Work
2.1 Multispectral Pedestrian Detection
2.2 Attention Mechanism
3 Proposed Method
3.1 Cross-Modality Feature Extraction Module
3.2 Spatial Attention Fusion Module
3.3 Channel Attention Fusion Module
4 Experiments
4.1 Dataset and Metric
4.2 Implementation Details
4.3 Quantitative Evaluation
4.4 Qualitative Evaluation
4.5 Ablation Study
5 Conclusion
References
Human Action Understanding

Overall-Distinctive GCN for Social Relation Recognition on Videos
1 Introduction
2 Related Work
3 Methodology
3.1 Framework Overview
3.2 Overall-level Character GCN
3.3 Distinctive-level Character GCN
3.4 Relation Classification
4 Experiments
4.1 Dataset
4.2 Implementation Details
4.3 Baseline Methods
4.4 Experiment Results
5 Ablation Study
6 Conculusion
References
Weakly-Supervised Temporal Action Localization with Regional Similarity Consistency
1 Introduction
2 Related Work
2.1 Weakly-supervised Temporal Action Localization