ECCV2022 &CVPR2022论文速递2022.7.19！

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 | XMem: 高质量长期视频分割！

效果超群！

标题：XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model

论文：https://arxiv.org/pdf/2207.07115.pdf

代码：https://github.com/hkchengrex/XMem

摘要：

我们提出了 XMem，这是一种用于长视频的视频对象分割架构，具有统一的特征内存存储，受 Atkinson-Shiffrin 内存模型的启发。先前关于视频对象分割的工作通常只使用一种类型的特征记忆。对于超过一分钟的视频，单个特征内存模型将内存消耗和准确性紧密联系在一起。相比之下，遵循 Atkinson-Shiffrin 模型，我们开发了一种架构，该架构包含多个独立但深度连接的特征记忆存储：快速更新的感觉记忆、高分辨率工作记忆和紧凑的持续长期记忆。至关重要的是，我们开发了一种记忆增强算法，该算法通常将积极使用的工作记忆元素整合到长期记忆中，从而避免记忆爆炸并最大限度地减少长期预测的性能衰减。结合新的内存读取机制，XMem 在长视频数据集上的性能大大超过了最先进的性能，同时在短视频上与最先进的方法（不适用于长视频）相当数据集。

最新论文整理

ECCV2022

Updated on : 19 Jul 2022

total number : 37

Rethinking Data Augmentation for Robust Visual Question Answering

• 论文/Paper: http://arxiv.org/pdf/2207.08739
• 代码/Code: https://github.com/ItemZheng/KDDAug

Semantic Novelty Detection via Relational Reasoning

• 论文/Paper: http://arxiv.org/pdf/2207.08699
• 代码/Code: None

Label2Label: A Language Modeling Framework for Multi-Attribute Learning

• 论文/Paper: http://arxiv.org/pdf/2207.08677
• 代码/Code: https://github.com/Li-Wanhua/Label2Label

Action-based Contrastive Learning for Trajectory Prediction

• 论文/Paper: http://arxiv.org/pdf/2207.08664
• 代码/Code: None

Towards High-Fidelity Single-view Holistic Reconstruction of Indoor Scenes

• 论文/Paper: http://arxiv.org/pdf/2207.08656
• 代码/Code: https://github.com/UncleMEDM/InstPIFu

FakeCLR: Exploring Contrastive Learning for Solving Latent Discontinuity in Data-Efficient GANs

• 论文/Paper: http://arxiv.org/pdf/2207.08630
• 代码/Code: https://github.com/iceli1007/FakeCLR.

Class-incremental Novel Class Discovery

• 论文/Paper: http://arxiv.org/pdf/2207.08605
• 代码/Code: https://github.com/OatmealLiu/class-iNCD

Dense Cross-Query-and-Support Attention Weighted Mask Aggregation for Few-Shot Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.08549
• 代码/Code: None

DID-M3D: Decoupling Instance Depth for Monocular 3D Object Detection

• 论文/Paper: http://arxiv.org/pdf/2207.08531
• 代码/Code: https://github.com/SPengLiang/DID-M3D.

Hierarchical Feature Alignment Network for Unsupervised Video Object Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.08485
• 代码/Code: https://github.com/NUST-Machine-Intelligence-Laboratory/HFAN

Open-world Semantic Segmentation via Contrasting and Clustering Vision-Language Embedding

• 论文/Paper: http://arxiv.org/pdf/2207.08455
• 代码/Code: None

TokenMix: Rethinking Image Mixing for Data Augmentation in Vision Transformers

• 论文/Paper: http://arxiv.org/pdf/2207.08409
• 代码/Code: https://github.com/Sense-X/TokenMix

MPIB: An MPI-Based Bokeh Rendering Framework for Realistic Partial Occlusion Effects

• 论文/Paper: http://arxiv.org/pdf/2207.08403
• 代码/Code: None

Adversarial Contrastive Learning via Asymmetric InfoNCE

• 论文/Paper: http://arxiv.org/pdf/2207.08374
• 代码/Code: https://github.com/yqy2001/A-InfoNCE

SepLUT: Separable Image-adaptive Lookup Tables for Real-time Image Enhancement

• 论文/Paper: http://arxiv.org/pdf/2207.08351
• 代码/Code: None

Learning with Recoverable Forgetting

• 论文/Paper: http://arxiv.org/pdf/2207.08224
• 代码/Code: None

Fast-MoCo: Boost Momentum-based Contrastive Learning with Combinatorial Patches

• 论文/Paper: http://arxiv.org/pdf/2207.08220
• 代码/Code: None

Zero-Shot Temporal Action Detection via Vision-Language Prompting

• 论文/Paper: http://arxiv.org/pdf/2207.08184
• 代码/Code: https://github.com/sauradip/STALE

Watermark Vaccine: Adversarial Attacks to Prevent Watermark Removal

• 论文/Paper: http://arxiv.org/pdf/2207.08178
• 代码/Code: None

FashionViL: Fashion-Focused Vision-and-Language Representation Learning

• 论文/Paper: http://arxiv.org/pdf/2207.08150
• 代码/Code: https://github.com/BrandonHanx/mmf.

E-NeRV: Expedite Neural Video Representation with Disentangled Spatial-Temporal Context

• 论文/Paper: http://arxiv.org/pdf/2207.08132
• 代码/Code: https://github.com/kyleleey/E-NeRV.

CATRE: Iterative Point Clouds Alignment for Category-level Object Pose Refinement

• 论文/Paper: http://arxiv.org/pdf/2207.08082
• 代码/Code: None

Neural Color Operators for Sequential Image Retouching

• 论文/Paper: http://arxiv.org/pdf/2207.08080
• 代码/Code: https://github.com/amberwangyili/neurop

Semi-Supervised Keypoint Detector and Descriptor for Retinal Image Matching

• 论文/Paper: http://arxiv.org/pdf/2207.07932
• 代码/Code: None

Learning Quality-aware Dynamic Memory for Video Object Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.07922
• 代码/Code: https://github.com/workforai/QDMN.

SPSN: Superpixel Prototype Sampling Network for RGB-D Salient Object Detection

• 论文/Paper: http://arxiv.org/pdf/2207.07898
• 代码/Code: https://github.com/Hydragon516/SPSN

JPerceiver: Joint Perception Network for Depth, Pose and Layout Estimation in Driving Scenes

• 论文/Paper: http://arxiv.org/pdf/2207.07895
• 代码/Code: at~\href{https://github.com/sunnyHelen/JPerceiver}{https://github.com/sunnyHelen/JPerceiver}.

You Should Look at All Objects

• 论文/Paper: http://arxiv.org/pdf/2207.07889
• 代码/Code: None

NeFSAC: Neurally Filtered Minimal Samples

• 论文/Paper: http://arxiv.org/pdf/2207.07872
• 代码/Code: https://github.com/cavalli1234/NeFSAC.

CLOSE: Curriculum Learning On the Sharing Extent Towards Better One-shot NAS

• 论文/Paper: http://arxiv.org/pdf/2207.07868
• 代码/Code: https://github.com/walkerning/aw_nas.

TS2-Net: Token Shift and Selection Transformer for Text-Video Retrieval

• 论文/Paper: http://arxiv.org/pdf/2207.07852
• 代码/Code: None

Cross-Domain Cross-Set Few-Shot Learning via Learning Compact and Aligned Representations

• 论文/Paper: http://arxiv.org/pdf/2207.07826
• 代码/Code: https://github.com/WentaoChen0813/CDCS-FSL

Bagging Regional Classification Activation Maps for Weakly Supervised Object Localization

• 论文/Paper: http://arxiv.org/pdf/2207.07818
• 代码/Code: https://github.com/zh460045050/BagCAMs.

Self-calibrating Photometric Stereo by Neural Inverse Rendering

• 论文/Paper: http://arxiv.org/pdf/2207.07815
• 代码/Code: https://github.com/junxuan-li/SCPS-NIR

Learning Long-Term Spatial-Temporal Graphs for Active Speaker Detection

• 论文/Paper: http://arxiv.org/pdf/2207.07783
• 代码/Code: https://github.com/SRA2/SPELL

Towards Understanding The Semidefinite Relaxations of Truncated Least-Squares in Robust Rotation Search

• 论文/Paper: http://arxiv.org/pdf/2207.08350
• 代码/Code: None

TransGrasp: Grasp Pose Estimation of a Category of Objects by Transferring Grasps from Only One Labeled Instance

• 论文/Paper: http://arxiv.org/pdf/2207.07861
• 代码/Code: https://github.com/yanjh97/TransGrasp.

CVPR2022

无