CVPR2022论文速递（2022.4.20）！共13篇！

整理：AI算法与图像处理

CVPR2022论文和代码整理：https://github.com/DWCTOD/CVPR2022-Papers-with-Code-Demo

欢迎关注公众号 AI算法与图像处理，获取更多干货：

大家好,  最近正在优化每周分享的CVPR论文, 目前考虑按照不同类别去分类,方便不同方向的小伙伴挑选自己感兴趣的论文哈

欢迎大家留言其他想法,  合适的话会采纳哈! 求个三连支持一波哈

Updated on : 20 Apr 2022

total number : 13

CVPR2022 Oral - 1 篇

Not All Tokens Are Equal: Human-centric Visual Analysis via Token Clustering Transformer  (Oral)

标题：并非所有token都是平等的：通过token聚类Transformer以人为际的视觉分析

• 论文/Paper: http://arxiv.org/pdf/2204.08680

• 代码/Code: https://github.com/zengwang430521/TCFormer

分类 / Classification - 1 篇

Revisiting Vicinal Risk Minimization for Partially Supervised Multi-Label Classification Under Data Scarcity

标题：根据数据稀缺，重新审视部分监督的多标签分类最小化

• 论文/Paper: http://arxiv.org/pdf/2204.08954

• 代码/Code: None

目标检测 / Object Detection - 2 篇

Modeling Missing Annotations for Incremental Learning in Object Detection

标题：对象检测中增量学习的建模缺失注释

• 论文/Paper: http://arxiv.org/pdf/2204.08766

• 代码/Code: https://github.com/fcdl94/MMA

Augmentation of Atmospheric Turbulence Effects on Thermal Adapted Object Detection Models

标题：增强热适应对象检测模型的大气湍流效应

• 论文/Paper: http://arxiv.org/pdf/2204.08745

• 代码/Code: None

语义分割/Segmentation - 2 篇

Dual-Domain Image Synthesis using Segmentation-Guided GAN

标题：双域图像合成使用分割引导GaN

• 论文/Paper: http://arxiv.org/pdf/2204.09015

• 代码/Code: https://github.com/denabazazian/Dual-Domain-Synthesis.

Proposal-free Lidar Panoptic Segmentation with Pillar-level Affinity

标题：免除柱级亲和力的免费激光雷达光学分割

• 论文/Paper: http://arxiv.org/pdf/2204.08744

• 代码/Code: None

GAN/ - 1 篇

Dual-Domain Image Synthesis using Segmentation-Guided GAN

标题：双域图像合成使用分割引导GaN

• 论文/Paper: http://arxiv.org/pdf/2204.09015

• 代码/Code: https://github.com/denabazazian/Dual-Domain-Synthesis.

超分/Super-Resolution - 2 篇

Self-Calibrated Efficient Transformer for Lightweight Super-Resolution

标题：用于轻质超分辨率的自校准高效Transformer

• 论文/Paper: http://arxiv.org/pdf/2204.08913

• 代码/Code: https://github.com/AlexZou14/SCET.

Edge-enhanced Feature Distillation Network for Efficient Super-Resolution

标题：高效超分辨率的边缘增强功能蒸馏网

• 论文/Paper: http://arxiv.org/pdf/2204.08759

• 代码/Code: https://github.com/icandle/EFDN.

Transformers - - 3 篇

Self-Calibrated Efficient Transformer for Lightweight Super-Resolution

标题：用于轻质超分辨率的自校准高效Transformer

• 论文/Paper: http://arxiv.org/pdf/2204.08913

• 代码/Code: https://github.com/AlexZou14/SCET.

Multimodal Token Fusion for Vision Transformers

标题：视觉Transformer的多模态token融合

• 论文/Paper: http://arxiv.org/pdf/2204.08721

• 代码/Code: None

Not All Tokens Are Equal: Human-centric Visual Analysis via Token Clustering Transformer  (Oral)

标题：并非所有token都是平等的：通过token聚类Transformer以人为际的视觉分析

• 论文/Paper: http://arxiv.org/pdf/2204.08680

• 代码/Code: https://github.com/zengwang430521/TCFormer

多模态 / Multimodal - 1 篇

Multimodal Token Fusion for Vision Transformers

标题：视觉Transformer的多模态token融合

• 论文/Paper: http://arxiv.org/pdf/2204.08721

• 代码/Code: None

其他/Other - 4 篇

Learning to Imagine: Diversify Memory for Incremental Learning using Unlabeled Data

标题：学习想象：使用未标记数据的增量学习多样化内存

• 论文/Paper: http://arxiv.org/pdf/2204.08932

• 代码/Code: https://github.com/TOM-tym/Learn-to-Imagine

An Efficient Domain-Incremental Learning Approach to Drive in All Weather Conditions

标题：在所有天气条件下驱动的有效域 - 增量学习方法

• 论文/Paper: http://arxiv.org/pdf/2204.08817

• 代码/Code: None

Incorporating Semi-Supervised and Positive-Unlabeled Learning for Boosting Full Reference Image Quality Assessment

标题：融入半监督和积极的未标记学习，用于提高完整参考图像质量评估

• 论文/Paper: http://arxiv.org/pdf/2204.08763

• 代码/Code: https://github.com/happycaoyue/JSPL

Self-Supervised Equivariant Learning for Oriented Keypoint Detection

标题：面向关键点检测的自我监督的等级学习

• 论文/Paper: http://arxiv.org/pdf/2204.08613

• 代码/Code: None