ECCV2022 &CVPR2022论文速递2022.7.15！

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 | CCPL：通用的风格迁移方法，艺术化、逼真的视频风格迁移！

标题：

CCPL: Contrastive Coherence Preserving Loss for Versatile Style Transfer

论文：https://arxiv.org/abs/2207.04808

代码：https://github.com/JarrentWu1031/CCPL

摘要：

在本文中，我们的目标是设计一种通用的风格迁移方法，该方法能够联合执行艺术、逼真和视频风格迁移，而无需在训练期间看到视频。以前的单帧方法假设对整个图像有很强的约束以保持时间一致性，这在许多情况下可能会被违反。相反，我们做了一个温和而合理的假设，即全局不一致性由局部不一致性支配，并设计了一种适用于局部补丁的通用对比相干保持损失 (CCPL)。CCPL 可以在风格转换期间保持内容源的连贯性，而不会降低风格化。此外，它拥有一个邻居调节机制，从而大大减少了局部失真并显着提高了视觉质量。除了在通用风格转换方面的卓越性能外，它还可以轻松扩展到其他任务，例如图像到图像的转换。此外，为了更好地融合内容和风格特征，我们提出了简单协方差变换（SCT）来有效地将内容特征的二阶统计与风格特征对齐。

最新论文整理

ECCV2022

Updated on : 15 Jul 2022

total number : 18

Bootstrapped Masked Autoencoders for Vision BERT Pretraining

• 论文/Paper: http://arxiv.org/pdf/2207.07116
• 代码/Code: https://github.com/lightdxy/bootmae

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

• 论文/Paper: http://arxiv.org/pdf/2207.07115
• 代码/Code: https://github.com/hkchengrex/XMem

Relighting4D: Neural Relightable Human from Videos

• 论文/Paper: http://arxiv.org/pdf/2207.07104
• 代码/Code: https://github.com/FrozenBurning/Relighting4D

ReAct: Temporal Action Detection with Relational Queries

• 论文/Paper: http://arxiv.org/pdf/2207.07097
• 代码/Code: https://github.com/sssste/react

Towards Grand Unification of Object Tracking

• 论文/Paper: http://arxiv.org/pdf/2207.07078
• 代码/Code: https://github.com/masterbin-iiau/unicorn

Semi-Supervised Temporal Action Detection with Proposal-Free Masking

• 论文/Paper: http://arxiv.org/pdf/2207.07059
• 代码/Code: https://github.com/sauradip/SPOT

Tree Structure-Aware Few-Shot Image Classification via Hierarchical Aggregation

• 论文/Paper: http://arxiv.org/pdf/2207.06989
• 代码/Code: https://github.com/remiMZ/HTS-ECCV22

ObjectBox: From Centers to Boxes for Anchor-Free Object Detection

• 论文/Paper: http://arxiv.org/pdf/2207.06985
• 代码/Code: https://github.com/mohsenzand/objectbox

Scene Text Recognition with Permuted Autoregressive Sequence Models

• 论文/Paper: http://arxiv.org/pdf/2207.06966
• 代码/Code: https://github.com/baudm/parseq

Learning Implicit Templates for Point-Based Clothed Human Modeling

• 论文/Paper: http://arxiv.org/pdf/2207.06955
• 代码/Code: https://github.com/jsnln/fite

Tackling Background Distraction in Video Object Segmentation

• 论文/Paper: http://arxiv.org/pdf/2207.06953
• 代码/Code: https://github.com/suhwan-cho/tbd

BayesCap: Bayesian Identity Cap for Calibrated Uncertainty in Frozen Neural Networks

• 论文/Paper: http://arxiv.org/pdf/2207.06873
• 代码/Code: https://github.com/explainableml/bayescap

Point-to-Box Network for Accurate Object Detection via Single Point Supervision

• 论文/Paper: http://arxiv.org/pdf/2207.06827
• 代码/Code: None

ConCL: Concept Contrastive Learning for Dense Prediction Pre-training in Pathology Images

• 论文/Paper: http://arxiv.org/pdf/2207.06733
• 代码/Code: https://github.com/tencentailabhealthcare/concl

Dynamic Low-Resolution Distillation for Cost-Efficient End-to-End Text Spotting

• 论文/Paper: http://arxiv.org/pdf/2207.06694
• 代码/Code: https://github.com/hikopensource/davar-lab-ocr

Temporal Action Detection with Global Segmentation Mask Learning

• 论文/Paper: http://arxiv.org/pdf/2207.06580
• 代码/Code: https://github.com/sauradip/TAGS

Supervised Attribute Information Removal and Reconstruction for Image Manipulation

• 论文/Paper: http://arxiv.org/pdf/2207.06555
• 代码/Code: https://github.com/nannanli999/airr

Lipschitz Continuity Retained Binary Neural Network

• 论文/Paper: http://arxiv.org/pdf/2207.06540
• 代码/Code: https://github.com/42shawn/lcr_bnn

CVPR2022

Updated on : 15 Jul 2022

total number : 1

Egocentric Scene Understanding via Multimodal Spatial Rectifier

• 论文/Paper: http://arxiv.org/pdf/2207.07077
• 代码/Code: https://github.com/tien-d/EgoDepthNormal