ICPR 2022

303 papers
The 2022 edition of the International Conference on Pattern Recognition (ICPR) featured 388 papers covering a broad spectrum of computer vision and machine learning topics. Key research directions included trajectory prediction using differentiable geometric constraints, learning safety boundaries for robotics, and discovering novel categories without predefined cluster counts. There was a focus on improving model robustness through automated shortcut discovery, counterfactual data augmentation, and selective consistency regularization for domain generalization. Methodological trends highlighted the integration of self-supervision with semi-supervised learning and the use of geometric self-supervision for semantic understanding. Specific applications explored during this edition included weather-aware object detection, cloth-changing person re-identification, and lightweight architectures for remote sensing land-cover classification. Prolific contributors to the proceedings included Xiaohua Xie, Jing Dong, Wei Wang, and Rita Cucchiara. The research collectively addressed challenges in out-of-distribution detection, low-rank matrix recovery, and probabilistic label aggregation for noisy datasets.

Top Research Directions

Research directions most represented by ICPR 2022 papers.

Semantic Alignment and Feature Mapping for Zero-Shot Learning 2
Probabilistic Trajectory Forecasting for Autonomous Systems 2
Modeling Multimodal Uncertainty in Video Prediction 2
Restoring Visual Structure in Low Light Environments 2
Accelerating Neural Architecture Search Without Full Model Training 2
Attention Gated Skip Connections for Selective Feature Fusion 2
Test Time Adaptation for Distribution Shift 2
Architectural Coupling of Positional Encodings in Transformers 1
Optimizing Image Models with Text Recognition Feedback 1
Learning Equivariant Representations through Predictable Transformations 1
Repurposing Pretrained Diffusion Models for Visual Perception 1
Continuous Orientation and Pose Estimation 1
Active Viewpoint Selection for 3D Reconstruction 1
Decomposing Facial Expressions into Action Units 1
Scaling Optimal Transport via One-Dimensional Projections 1

Top Authors

AuthorPapers
Jing Dong4
Xiaohua Xie4
Didier Stricker3
Lorenzo Baraldi3
Marcella Cornia3
Mubarak Shah3
Ngai Wong3
Ngan Le3
Ram Nevatia3
Rita Cucchiara3
Wei Wang3
Ajmal Saeed Mian2
Alberto Del Bimbo2
Andreas Dengel2
Aymeric Histace2
Bo Peng2
Buddhika Laknath Semage2
Chenglong Li2
Christopher Zach2
Chuan-Sheng Foo2

Impactful Papers

Papers from ICPR 2022 that shaped research clusters.

Foundational

Influential

Papers

Title
Shuffle & Divide: Contrastive Learning for Long Text
Simple Yet Effective Approach to Repetitive Behavior Classification based on Siamese Network
Space-correlated Contrastive Representation Learning with Multiple Instances
Speckle Image Restoration without Clean Data
Speeding up Heterogeneous Federated Learning with Sequentially Trained Superclients
Spot the Difference: A Novel Task for Embodied Agents in Changing Environments
SSL-DC: Improving Transductive Few-Shot Learning via Self-Supervised Learning and Distribution Calibration
StyleVC: Non-Parallel Voice Conversion with Adversarial Style Generalization
Swapping Semantic Contents for Mixing Images
Synthesis in Style: Semantic Segmentation of Historical Documents using Synthetic Data
Tag-Based Attention Guided Bottom-Up Approach for Video Instance Segmentation
Temporal Alignment for History Representation in Reinforcement Learning
Temporal Discriminative Micro-Expression Recognition via Graph Contrastive Learning
Temporal Shift and Attention Modules for Graphical Skeleton Action Recognition
TextSRNet: Scene Text Super-Resolution Based on Contour Prior and Atrous Convolution
The First Challenge on Moving Object Detection and Tracking in Satellite Videos: Methods and Results
The LAM Dataset: A Novel Benchmark for Line-Level Handwritten Text Recognition
Towards Automatic Parsing of Structured Visual Content through the Use of Synthetic Data
Towards Improved Skin Lesion Classification using Metadata Supervision
Towards Interpretable Deep Reinforcement Learning Models via Inverse Reinforcement Learning
Towards Positive Jacobian: Learn to Postprocess for Diffeomorphic Image Registration with Matrix Exponential
Towards Robust Community Detection via Extreme Adversarial Attacks
Towards Scalable and Efficient Client Selection for Federated Object Detection
Towards Streaming Egocentric Action Anticipation
Transformer-based Scene Graph Generation Network With Relational Attention Module
Transformers for 1D signals in Parkinson's disease detection from gait
Transforming Image Generation from Scene Graphs
TwinLSTM: Two-channel LSTM Network for Online Action Detection
U-Former: Improving Monaural Speech Enhancement with Multi-head Self and Cross Attention
Uncertain Label Correction via Auxiliary Action Unit Graphs for Facial Expression Recognition
Uncertainty Aware System Identification with Universal Policies
Unified Recurrence Modeling for Video Action Anticipation
Unsteady Flow Field Prediction: Understanding the Dynamics in an Intuitive Physics Way
Unsupervised Domain Adaptation for Video Transformers in Action Recognition
Unsupervised Hierarchical Disentanglement for Video Prediction
Unsupervised Low-light Image Enhancement with Decoupled Networks
Variance of Local Contribution: an Unsupervised Image Quality Assessment for Face Recognition
VASAD: a Volume and Semantic dataset for Building Reconstruction from Point Clouds
Video Generation from Text Employing Latent Path Construction for Temporal Modeling
VideoPipe 2022 Challenge: Real-World Video Understanding for Urban Pipe Inspection
VidHarm: A Clip Based Dataset for Harmful Content Detection
Virtual Tasks but Real Gains: Improving Multi-Task Learning
Vision Transformers Based Classification for Glaucomatous Eye Condition
Visual Transformers with Primal Object Queries for Multi-Label Image Classification
VPTR: Efficient Transformers for Video Prediction
VReBERT: A Simple and Flexible Transformer for Visual Relationship Detection
Weakly Supervised Object Localization Using Long-Range Semantic Foreground Activation
Weak-supervised Visual Geo-localization via Attention-based Knowledge Distillation
WeStcoin: Weakly-Supervised Contextualized Text Classification with Imbalance and Noisy Labels
What makes you, you? Analyzing Recognition by Swapping Face Parts

« prev Page 6/7 next »