ECCV2022论文汇总:检测/分割/跟踪/3D/深度估计/姿态解算等多个方向!

作者 | 汽车人

编辑 | Autobox

目前,公众号正向大家广泛征稿中,欢迎童鞋们投稿,我们将有一定的稿费支持哦,详细信息请点击: 汽车人,快来投稿了!

数据集

COO: Comic Onomatopoeia Dataset for Recognizing Arbitrary or Truncated Texts

地址:https://arxiv.org/pdf/2207.04675

Github:https://github.com/ku21fan/COO-Comic-Onomatopoeia

视觉transformer

[1].k-means Mask Transformer

论文链接: http://arxiv.org/pdf/2207.04044

代码链接: https://github.com/google-research/deeplab2

[2].Weakly Supervised Grounding for VQA in Vision-Language Transformers

论文链接: http://arxiv.org/pdf/2207.02334

代码链接: https://github.com/aurooj/wsg-vqa-vltransformers

[3].Wave-ViT: Unifying Wavelet and Transformers for Visual Representation Learning

论文链接: http://arxiv.org/pdf/2207.04978

代码链接: https://github.com/YehLi/ImageNetModel

[4].CoMER: Modeling Coverage for Transformer-based Handwritten Mathematical Expression Recognition

论文链接: http://arxiv.org/pdf/2207.04410

代码链接: https://github.com/Green-Wood/CoMER

[5].MaxViT: Multi-Axis Vision Transformer

论文介绍了一种高效且可扩展的注意力模型,称之为多轴注意力

论文链接:MaxViT: Multi-Axis Vision Transformer

[6].V2X-ViT: Vehicle-to-Everything Cooperative Perception with Vision Transformer.

主页链接:https://github.com/DerrickXuNu/v2x-vit

目标检测

[1].ObjectBox: From Centers to Boxes for Anchor-Free Object Detection

Anchor-free模型新范式;

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

代码链接:https://github.com/MohsenZand/ObjectBox

[2].Dense Teacher: Dense Pseudo-Labels for Semi-supervised Object Detection

半监督目标检测方法:Dense Teacher

论文链接: http://arxiv.org/pdf/2207.02541

[3]Should All Proposals be Treated Equally in Object Detection?

目标检测中的采样讨论!

论文链接: http://arxiv.org/pdf/2207.03520

[4].ViTDet: Exploring Plain Vision Transformer Backbones for Object Detection.

来自何凯明团队,证明了将普通的、非分层的视觉 Transformer 作为主干网络进行目标检测的可行性;

论文链接:Exploring Plain Vision Transformer Backbones for Object Detection

代码链接:https://github.com/facebookresearch/detectron2/tree/main/projects/ViTDet

目标跟踪

[1].Towards Grand Unification of Object Tracking 多任务统一框架Unicorn,解决四个跟踪问题(SOT、MOT、VOS、MOTS)。Unicorn 在 8 个跟踪数据集(包括 LaSOT、TrackingNet、MOT17、BDD100K、DAVIS16-17、MOTS20 和 BDD100K MOTS)中的表现与其特定任务的counterparts相当或更好。

论文链接:Towards Grand Unification of Object Tracking

代码链接:https://github.com/MasterBin-IIAU/Unicorn

[2].MOTR: End-to-End Multiple-Object Tracking with TRansformer

论文链接:https://arxiv.org/pdf/2105.03247.pdf

代码链接:https://github.com/megvii-research/MOTR

3D检测

[1].FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection

全卷积anchor-free方案3D目标检测;

论文链接:FCAF3D: Fully Convolutional Anchor-Free 3D Object Detection

代码链接:https://github.com/SamsungLabs/fcaf3d

[2].PETR: Position Embedding Transformation for Multi-View 3D Object Detection

论文链接:PETR: Position Embedding Transformation for Multi-View 3D Object Detection

代码链接:https://github.com/megvii-research/PETR

[3].AutoAlign: Pixel-Instance Feature Aggregation for Multi-Modal 3D Object Detection

多模态3D目标检测

[1].论文链接:AutoAlign: Pixel-Instance Feature Aggregation for Multi-Modal 3D Object Detection

语义分割

[1].Domain Adaptive Video Segmentation via Temporal Pseudo Supervision

论文链接: http://arxiv.org/pdf/2207.02372

代码链接: https://github.com/xing0047/tps

[2].OSFormer: One-Stage Camouflaged Instance Segmentation with Transformers

论文链接: http://arxiv.org/pdf/2207.02255

代码链接: https://github.com/pjlallen/osformer

[3].Prototypical Contrast Adaptation for Domain Adaptive Semantic Segmentation.

无监督域适应方法

论文链接:https://arxiv.org/pdf/2207.06654.pdf

代码链接:https://github.com/jiangzhengkai/ProCA

[4].2DPASS: 2D Priors Assisted Semantic Segmentation on LiDAR Point Clouds 2D先验辅助的激光雷达点云语义分割

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

代码链接:https://github.com/yanx27/2dpass

车道线检测

[1].PersFormer: a New Baseline for 3D Laneline Detection.

3D车道线检测新基线;

论文链接:PersFormer: 3D Lane Detection via Perspective Transformer and the OpenLane Benchmark

代码链接:https://github.com/OpenPerceptionX/PersFormer_3DLane

点云相关

[1].Open-world Semantic Segmentation for LIDAR Point Clouds

论文链接: http://arxiv.org/pdf/2207.01452

代码链接: 

https://github.com/jun-cen/open_world_3d_semantic_segmentation

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

论文链接: http://arxiv.org/pdf/2207.04397

深度估计

[1].Physical Attack on Monocular Depth Estimation with Optimal Adversarial Patches

论文链接: http://arxiv.org/pdf/2207.04718

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

论文链接: http://arxiv.org/pdf/2207.04680

6D姿态估计

[1].Gen6D: Generalizable Model-Free 6-DoF Object Pose Estimation from RGB Images.

论文链接:Gen6D: Generalizable Model-Free 6-DoF Object Pose Estimation from RGB Images

代码链接:Gen6D: Generalizable Model-Free 6-DoF Object Pose Estimation from RGB Images

其它

[1].MVSTER: Epipolar Transformer for Efficient Multi-View Stereo.

论文链接:MVSTER: Epipolar Transformer for Efficient Multi-View Stereo

代码链接:https://github.com/JeffWang987/MVSTER

Workshop竞赛汇总

[1].Visual Inductive Priors for Data-Efficient Computer Vision

链接:https://vipriors.github.io/challenges/

[2].ECCV'22 ChaLearn Seasons in Drift Challenge 

链接:https://codalab.lisn.upsaclay.fr/competitions/4272

[3].ECCV'22 ChaLearn Seasons in Drift Challenge

链接:https://codalab.lisn.upsaclay.fr/competitions/4273

[4].ECCV'22 ChaLearn Seasons in Drift Challenge 

链接:https://codalab.lisn.upsaclay.fr/competitions/4276

[5].ECCV'22 ChaLearn Sign Spotting Challenge Challenge 

链接:https://codalab.lisn.upsaclay.fr/competitions/4198

[6].ECCV'22 ChaLearn Sign Spotting Challenge Challenge 

链接:https://codalab.lisn.upsaclay.fr/competitions/4199

[7].ECCV DeeperAction Challenge - MultiSports Track on Human Action Detection

链接:https://codalab.lisn.upsaclay.fr/competitions/3736

[8].ECCV DeeperAction Challenge - FineAction Track on Temporal Action Localization

链接:https://codalab.lisn.upsaclay.fr/competitions/4386

[9].ECCV DeeperAction Challenge - UrbanPipe Track on Fine-grained Video Anomaly Recognition.

链接:https://codalab.lisn.upsaclay.fr/competitions/4303

[10].ECCV 2022 WCPA Challenge: From Face, Body and Fashion to 3D Virtual Avatars Ⅰ 

链接:

https://tianchi.aliyun.com/competition/entrance/531958/introduction?spm=5176.12281957.1004.3.2fe13eafY89Q32

[11].ECCV 2022 WCPA Challenge: From Face, Body and Fashion to 3D Virtual Avatars Ⅱ

链接:

https://tianchi.aliyun.com/competition/entrance/531961/introduction?spm=5176.12281957.1004.2.2fe13eafY89Q32

自动驾驶之心】全栈技术交流群

自动驾驶之心是国内首个自动驾驶开发者社区,聚焦目标检测、语义分割、关键点检测、车道线、目标跟踪、3D感知、多传感器融合、SLAM、高精地图、规划控制、AI模型部署落地等方向;

加入我们:自动驾驶之心技术交流群汇总!

自动驾驶之心【知识星球】

想要了解更多自动驾驶感知(分类、检测、分割、关键点、车道线、3D感知、多传感器融合、目标跟踪)、自动驾驶定位建图(SLAM、高精地图)、自动驾驶规划控制、领域技术方案、AI模型部署落地实战、行业动态、岗位发布,欢迎扫描下方二维码,加入自动驾驶之心知识星球,这里汇聚行业和学术界大佬,前沿技术方向尽在掌握中,期待交流!

你可能感兴趣的:(人工智能,计算机视觉,深度学习,github,机器学习)