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：

SIGGRAPH 2022 | Honorable Mentions 稀疏椭偏法，重建效果接近真实物体！已开源

主页：http://vclab.kaist.ac.kr/siggraph2022p1/index.html 代码：https://github.com/KAIST-VCLAB/SparseEllipsometry

摘要：

提出了稀疏椭偏法，这是一种便携式偏振采集方法，可捕获极化 SVBRDF 和 3D 形状同时进行。我们的手持设备由现成的固定光学元件组成。每个对象的总采集时间不是几天，而是二十到三十分钟。我们开发了一个完整的偏振 SVBRDF 模型，包括漫反射和镜面反射分量，以及单次散射，并设计了一种新的偏振逆渲染算法，通过生成建模对镜面反射样本进行数据增强。我们的结果与最近捕获的真实世界对象的极化 BRDF 的grouth truth 数据集非常一致。

最新论文整理

ECCV2022

Updated on : 12 Jul 2022

total number : 18

语义分割/Segmentation - 2 篇

Personalizing Federated Medical Image Segmentation via Local Calibration

标题：通过本地校准个性化联合医学图像分割

• 论文/Paper: http://arxiv.org/pdf/2207.04655
• 代码/Code: https://github.com/jcwang123/FedLC

2DPASS: 2D Priors Assisted Semantic Segmentation on LiDAR Point Clouds

标题：2DPASS：2D先验辅助语义分割在LiDAR Point Clouds上

• 论文/Paper: http://arxiv.org/pdf/2207.04397
• 代码/Code: None

Transformers - - 2 篇

Wave-ViT: Unifying Wavelet and Transformers for Visual Representation Learning

标题：Wave-ViT：统一的小波和Transformers用于视觉表示学习

• 论文/Paper: http://arxiv.org/pdf/2207.04978
• 代码/Code: https://github.com/YehLi/ImageNetModel

CoMER: Modeling Coverage for Transformer-based Handwritten Mathematical Expression Recognition

标题：CoMER：基于Transformer的手写数学表达识别的建模覆盖范围

• 论文/Paper: http://arxiv.org/pdf/2207.04410
• 代码/Code: https://github.com/Green-Wood/CoMER

点云/Point Clouds - 1 篇

2DPASS: 2D Priors Assisted Semantic Segmentation on LiDAR Point Clouds

标题：2DPASS：2D先验辅助语义分割在LiDar Point Clouds上

• 论文/Paper: http://arxiv.org/pdf/2207.04397
• 代码/Code: None

深度估计/Depth Estimation - 2 篇

Physical Attack on Monocular Depth Estimation with Optimal Adversarial Patches

标题：最佳对抗斑块对单目深度估计的物理攻击

• 论文/Paper: http://arxiv.org/pdf/2207.04718
• 代码/Code: None

Towards Scale-Aware, Robust, and Generalizable Unsupervised Monocular Depth Estimation by Integrating IMU Motion Dynamics

标题：通过整合IMU运动动力学，尺度感知，鲁棒和可概括的无监督单目估计

• 论文/Paper: http://arxiv.org/pdf/2207.04680
• 代码/Code: https://github.com/SenZHANG-GitHub/ekf-imu-depth

其他/Other - 12 篇

Demystifying Unsupervised Semantic Correspondence Estimation

标题：揭开无监督的语义对应估计

• 论文/Paper: http://arxiv.org/pdf/2207.05054
• 代码/Code: None

Fast-Vid2Vid: Spatial-Temporal Compression for Video-to-Video Synthesis

标题：Fast-VID2VID：视频与视频合成的时空压缩

• 论文/Paper: http://arxiv.org/pdf/2207.05049
• 代码/Code: https://github.com/fast-vid2vid/fast-vid2vid

A Closer Look at Invariances in Self-supervised Pre-training for 3D Vision

标题：仔细观察3D视觉的自监督预训练的不变性

• 论文/Paper: http://arxiv.org/pdf/2207.04997
• 代码/Code: None

Cross-modal Prototype Driven Network for Radiology Report Generation

标题：用于放射学报告生成的跨模式原型驱动网络

• 论文/Paper: http://arxiv.org/pdf/2207.04818
• 代码/Code: None

CCPL: Contrastive Coherence Preserving Loss for Versatile Style Transfer

标题：CCPL：多功能风格转移的对比度连贯性损失

• 论文/Paper: http://arxiv.org/pdf/2207.04808
• 代码/Code: https://github.com/JarrentWu1031/CCPL

DCCF: Deep Comprehensible Color Filter Learning Framework for High-Resolution Image Harmonization

标题：DCCF：高分辨率图像协调的深度可理解的颜色滤波器学习框架

• 论文/Paper: http://arxiv.org/pdf/2207.04788
• 代码/Code: None

Geometry-aware Single-image Full-body Human Relighting

标题：几何学意识到的单张全身人的重新Relighting

• 论文/Paper: http://arxiv.org/pdf/2207.04750
• 代码/Code: None

COO: Comic Onomatopoeia Dataset for Recognizing Arbitrary or Truncated Texts

标题：COO：漫画拟言数据集用于识别任意文本或截断的文本

• 论文/Paper: http://arxiv.org/pdf/2207.04675
• 代码/Code: https://github.com/ku21fan/COO-Comic-Onomatopoeia

Hierarchical Latent Structure for Multi-Modal Vehicle Trajectory Forecasting

标题：多模式车辆轨迹预测的分层潜在结构

• 论文/Paper: http://arxiv.org/pdf/2207.04624
• 代码/Code: https://github.com/d1024choi/HLSTrajForecast

Batch-efficient EigenDecomposition for Small and Medium Matrices

标题：中小型矩阵的批处理特征分类

• 论文/Paper: http://arxiv.org/pdf/2207.04228
• 代码/Code: None

Few 'Zero Level Set'-Shot Learning of Shape Signed Distance Functions in Feature Space

标题：在特征空间中，很少有“零级设置”的形状学习签名距离功能

• 论文/Paper: http://arxiv.org/pdf/2207.04161
• 代码/Code: None

Audio-Visual Segmentation

标题：视听细分

• 论文/Paper: http://arxiv.org/pdf/2207.05042
• 代码/Code: https://github.com/OpenNLPLab/AVSBench

CVPR2022

Updated on : 12 Jul 2022

total number : 1

Instance Shadow Detection with A Single-Stage Detector

• 论文/Paper: http://arxiv.org/pdf/2207.04614
• 代码/Code: None