CVPR2022论文速递（2022.4.19）！共19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

Updated on : 19 Apr 2022

total number : 19

CVPR2022 Oral - 2 篇

Detecting Deepfakes with Self-Blended Images (Oral)

标题：用自混合图像检测Deepfakes

• 论文/Paper: http://arxiv.org/pdf/2204.08376
• 代码/Code: https://github.com/mapooon/SelfBlendedImages

Animal Kingdom: A Large and Diverse Dataset for Animal Behavior Understanding (Oral)

标题：Animal Kingdom：用于动物行为理解的大规模多样化的数据集

• 论文/Paper: http://arxiv.org/pdf/2204.08129
• 代码/Code: None

目标检测 / Object Detection - 2 篇

Entropy-based Active Learning for Object Detection with Progressive Diversity Constraint

标题：基于熵的主动学习，用于具有渐进分集约束的目标检测

• 论文/Paper: http://arxiv.org/pdf/2204.07965
• 代码/Code: None

Target-Relevant Knowledge Preservation for Multi-Source Domain Adaptive Object Detection

标题：目标相关知识保存用于针对多源域自适应目标检测

• 论文/Paper: http://arxiv.org/pdf/2204.07964
• 代码/Code: None

目标跟踪/Object Tracking - 1 篇

Detecting, Tracking and Counting Motorcycle Rider Traffic Violations on Unconstrained Roads

标题：无约束的道路上检测，跟踪和计数摩托车骑手交通违规行为

• 论文/Paper: http://arxiv.org/pdf/2204.08364
• 代码/Code: None

语义分割/Segmentation - 2 篇

Temporally Efficient Vision Transformer for Video Instance Segmentation

标题：用于视频实例分段的时间高效视觉Transformer

• 论文/Paper: http://arxiv.org/pdf/2204.08412
• 代码/Code: https://github.com/hustvl/TeViT.

Augmentation Invariance and Adaptive Sampling in Semantic Segmentation of Agricultural Aerial Images

标题：农业空中图像语义分割中的增强不变性和自适应抽样

• 论文/Paper: http://arxiv.org/pdf/2204.07969
• 代码/Code: None

超分/Super-Resolution - 2 篇

Fast and Memory-Efficient Network Towards Efficient Image Super-Resolution

标题：快速和记忆高效的网络，高效图像超分辨率

• 论文/Paper: http://arxiv.org/pdf/2204.08397
• 代码/Code: https://github.com/NJU-Jet/FMEN.

Unsupervised domain adaptation and super resolution on drone images for autonomous dry herbage biomass estimation

标题：无人驾驶的域适应与自主干草生物量估计的无人机图像的超分辨率

• 论文/Paper: http://arxiv.org/pdf/2204.08271
• 代码/Code: None

Transformers - - 3 篇

Temporally Efficient Vision Transformer for Video Instance Segmentation

标题：用于视频实例分段的时间高效视觉Transformer

• 论文/Paper: http://arxiv.org/pdf/2204.08412
• 代码/Code: https://github.com/hustvl/TeViT.

Safe Self-Refinement for Transformer-based Domain Adaptation

标题：基于Transformer的域适应的安全自细化

• 论文/Paper: http://arxiv.org/pdf/2204.07683
• 代码/Code: None

Multi-Frame Self-Supervised Depth with Transformers

标题：Transformer的多帧自监督深度

• 论文/Paper: http://arxiv.org/pdf/2204.07616
• 代码/Code: None

其他/Other - 8 篇

Deep Equilibrium Optical Flow Estimation

标题：深度平衡光流量估计

• 论文/Paper: http://arxiv.org/pdf/2204.08442
• 代码/Code: None

OMG: Observe Multiple Granularities for Natural Language-Based Vehicle Retrieval

标题：OMG：观察基于语言的车辆检索多个粒度

• 论文/Paper: http://arxiv.org/pdf/2204.08209
• 代码/Code: https://github.com/dyhBUPT/OMG.

Video Action Detection: Analysing Limitations and Challenges

标题：视频动作检测：分析局限性和挑战

• 论文/Paper: http://arxiv.org/pdf/2204.07892
• 代码/Code: None

Towards a Deeper Understanding of Skeleton-based Gait Recognition

标题：朝着更深入地了解基于骨架的步态识别

• 论文/Paper: http://arxiv.org/pdf/2204.07855
• 代码/Code: None

Interactiveness Field in Human-Object Interactions

标题：人-对象交互中的互动场

• 论文/Paper: http://arxiv.org/pdf/2204.07718
• 代码/Code: https://github.com/Foruck/Interactiveness-Field.

Pushing the Performance Limit of Scene Text Recognizer without Human Annotation

标题：推动场景文本识别器的性能限制，没有人为注释

• 论文/Paper: http://arxiv.org/pdf/2204.07714
• 代码/Code: None

It is Okay to Not Be Okay: Overcoming Emotional Bias in Affective Image Captioning by Contrastive Data Collection

标题：不是没问题：通过对比数据收集克服情感形象标题中的情感偏见

• 论文/Paper: http://arxiv.org/pdf/2204.07660
• 代码/Code: None

Deep Unlearning via Randomized Conditionally Independent Hessians

标题：通过随机条件独立的Hessians深度无光

• 论文/Paper: http://arxiv.org/pdf/2204.07655
• 代码/Code: https://github.com/vsingh-group/LCODEC-deep-unlearning