ECCV2022 &CVPR2022论文速递2022.7.26！

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

ECCV2022论文和代码整理：https://github.com/DWCTOD/ECCV2022-Papers-with-Code-Demo

最新成果demo展示：

ECCV2022 | AI感知背景和选框合成最合适的图片

论文：https://arxiv.org/pdf/2204.00125.pdf ECCV2022 汇总：https://github.com/DWCTOD/ECCV2022-Papers-with-Code-Demo/

摘要：合成感知对象搜索旨在在给定背景图像和查询边界框的情况下找到最兼容的合成对象。以前的工作侧重于学习前景对象和背景之间的兼容性，但未能从大规模数据中学习其他重要因素，即几何和照明。为了更进一步，本文提出了 GALA（Geometry-and-Lighting-Aware），这是一种通用的前景对象搜索方法，对开放世界图像合成的几何和照明兼容性进行判别建模。值得注意的是，它在 CAIS 数据集上取得了最先进的结果，并在大规模开放世界数据集（即，Pixabay 和 Open Images）上很好地推广。此外，我们的方法可以有效地处理非框场景，即用户只提供背景图像而没有任何输入边界框。构建了一个网络演示（参见补充材料），以展示所提出的方法在前景对象的合成感知搜索和自动位置/比例预测方面的应用。

最新论文整理

ECCV2022

Updated on : 26 Jul 2022

total number : 36

Dynamic 3D Scene Analysis by Point Cloud Accumulation

• 论文/Paper: http://arxiv.org/pdf/2207.12394
• 代码/Code: None

CelebV-HQ: A Large-Scale Video Facial Attributes Dataset

• 论文/Paper: http://arxiv.org/pdf/2207.12393
• 代码/Code: https://github.com/CelebV-HQ/CelebV-HQ

MemSAC: Memory Augmented Sample Consistency for Large Scale Domain Adaptation

• 论文/Paper: http://arxiv.org/pdf/2207.12389
• 代码/Code: None

Deforming Radiance Fields with Cages

• 论文/Paper: http://arxiv.org/pdf/2207.12298
• 代码/Code: None

Equivariance and Invariance Inductive Bias for Learning from Insufficient Data

• 论文/Paper: http://arxiv.org/pdf/2207.12258
• 代码/Code: https://github.com/Wangt-CN/EqInv

Active Learning Strategies for Weakly-supervised Object Detection

• 论文/Paper: http://arxiv.org/pdf/2207.12112
• 代码/Code: https://github.com/huyvvo/BiB.

Black-box Few-shot Knowledge Distillation

• 论文/Paper: http://arxiv.org/pdf/2207.12106
• 代码/Code: https://github.com/nphdang/FS-BBT

W2N:Switching From Weak Supervision to Noisy Supervision for Object Detection

• 论文/Paper: http://arxiv.org/pdf/2207.12104
• 代码/Code: https://github.com/1170300714/w2n_wsod.

IGFormer: Interaction Graph Transformer for Skeleton-based Human Interaction Recognition

• 论文/Paper: http://arxiv.org/pdf/2207.12100
• 代码/Code: None

Balancing Stability and Plasticity through Advanced Null Space in Continual Learning

• 论文/Paper: http://arxiv.org/pdf/2207.12061
• 代码/Code: None

3D Siamese Transformer Network for Single Object Tracking on Point Clouds

• 论文/Paper: http://arxiv.org/pdf/2207.11995
• 代码/Code: None

RA-Depth: Resolution Adaptive Self-Supervised Monocular Depth Estimation

• 论文/Paper: http://arxiv.org/pdf/2207.11984
• 代码/Code: None

Reference-based Image Super-Resolution with Deformable Attention Transformer

• 论文/Paper: http://arxiv.org/pdf/2207.11938
• 代码/Code: None

Optimal Boxes: Boosting End-to-End Scene Text Recognition by Adjusting Annotated Bounding Boxes via Reinforcement Learning

• 论文/Paper: http://arxiv.org/pdf/2207.11934
• 代码/Code: None

NeuMesh: Learning Disentangled Neural Mesh-based Implicit Field for Geometry and Texture Editing

• 论文/Paper: http://arxiv.org/pdf/2207.11911
• 代码/Code: None

Domain Adaptive Person Search

• 论文/Paper: http://arxiv.org/pdf/2207.11898
• 代码/Code: https://github.com/caposerenity/DAPS.

On Mitigating Hard Clusters for Face Clustering

• 论文/Paper: http://arxiv.org/pdf/2207.11895
• 代码/Code: https://github.com/echoanran/On-Mitigating-Hard-Clusters.

Salient Object Detection for Point Clouds

• 论文/Paper: http://arxiv.org/pdf/2207.11889
• 代码/Code: None

VizWiz-FewShot: Locating Objects in Images Taken by People With Visual Impairments

• 论文/Paper: http://arxiv.org/pdf/2207.11810
• 代码/Code: None

Weakly-Supervised Temporal Action Detection for Fine-Grained Videos with Hierarchical Atomic Actions

• 论文/Paper: http://arxiv.org/pdf/2207.11805
• 代码/Code: None

Cross-Modal 3D Shape Generation and Manipulation

• 论文/Paper: http://arxiv.org/pdf/2207.11795
• 代码/Code: None

Learning Dynamic Facial Radiance Fields for Few-Shot Talking Head Synthesis

• 论文/Paper: http://arxiv.org/pdf/2207.11770
• 代码/Code: None

Label-Guided Auxiliary Training Improves 3D Object Detector

• 论文/Paper: http://arxiv.org/pdf/2207.11753
• 代码/Code: None

Combining Internal and External Constraints for Unrolling Shutter in Videos

• 论文/Paper: http://arxiv.org/pdf/2207.11725
• 代码/Code: None

TIPS: Text-Induced Pose Synthesis

• 论文/Paper: http://arxiv.org/pdf/2207.11718
• 代码/Code: None

Improving Test-Time Adaptation via Shift-agnostic Weight Regularization and Nearest Source Prototypes

• 论文/Paper: http://arxiv.org/pdf/2207.11707
• 代码/Code: None

Learning Graph Neural Networks for Image Style Transfer

• 论文/Paper: http://arxiv.org/pdf/2207.11681
• 代码/Code: None

Self-Support Few-Shot Semantic Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.11549
• 代码/Code: \url{https://github.com/fanq15/SSP}.

Contrastive Monotonic Pixel-Level Modulation

• 论文/Paper: http://arxiv.org/pdf/2207.11517
• 代码/Code: https://github.com/lukun199/MonoPix.

Active Pointly-Supervised Instance Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.11493
• 代码/Code: None

CompNVS: Novel View Synthesis with Scene Completion

• 论文/Paper: http://arxiv.org/pdf/2207.11467
• 代码/Code: None

When Counting Meets HMER: Counting-Aware Network for Handwritten Mathematical Expression Recognition

• 论文/Paper: http://arxiv.org/pdf/2207.11463
• 代码/Code: https://github.com/LBH1024/CAN.

UC-OWOD: Unknown-Classified Open World Object Detection

• 论文/Paper: http://arxiv.org/pdf/2207.11455
• 代码/Code: https://github.com/JohnWuzh/UC-OWOD.

Meta Spatio-Temporal Debiasing for Video Scene Graph Generation

• 论文/Paper: http://arxiv.org/pdf/2207.11441
• 代码/Code: None

PS-NeRF: Neural Inverse Rendering for Multi-view Photometric Stereo

• 论文/Paper: http://arxiv.org/pdf/2207.11406
• 代码/Code: None

Neural-Sim: Learning to Generate Training Data with NeRF

• 论文/Paper: http://arxiv.org/pdf/2207.11368
• 代码/Code: None

CVPR2022

Updated on : 26 Jul 2022

total number : 4

Revisiting AP Loss for Dense Object Detection: Adaptive Ranking Pair Selection

• 论文/Paper: http://arxiv.org/pdf/2207.12042
• 代码/Code: https://github.com/Xudangliatiger/APE-Loss.

Behind Every Domain There is a Shift: Adapting Distortion-aware Vision Transformers for Panoramic Semantic Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.11860
• 代码/Code: https://github.com/jamycheung/Trans4PASS

Audio-driven Neural Gesture Reenactment with Video Motion Graphs

• 论文/Paper: http://arxiv.org/pdf/2207.11524
• 代码/Code: None

PieTrack: An MOT solution based on synthetic data training and self-supervised domain adaptation

• 论文/Paper: http://arxiv.org/pdf/2207.11325
• 代码/Code: None