H²O 行人交互检测数据集

ImagesH²O is composed of the 10 301 images fromV-COCOdataset to which are added 3 635 images which mostly contain interactions between people.Extra-images sources:MS-COCOdatasetHuman-Interaction-ImagesdatasetBIT-InteractiondatasetSBU-Kinect-InteractiondatasetTV-Human-interactiondatasetPascal-VocdatasetShakeFive2datasetUT-InteractiondatasetWeb imagesAnnotationsAll H²O images have been annotated with a new taxonomy of verbs including human-to-object and human-to-human interactions.This taxonomy is composed of 51 verbs divided into 5 categories:Verbs decribing the general posture of the subjectVerbs related to the way the subject is movingVerbs used for interactions with objectVerbs describing human-to-human interactionsVerbs of interactions involving strength or violenceCitationA. Orcesi, R. Audigier, F. Poka Toukam and B. Luvison, “Detecting Human-to-Human-or-Object (H2O) Interactions with DIABOLO",2021 IEEE International Conference on Automatic Face and Gesture Recognition (FG2021),https://arxiv.org/abs/2201.02396@INPROCEEDINGS{h2o_dataset_2021, author={Orcesi, Astrid and Audigier, Romaric and Poka Toukam, Fritz and Luvison, Bertrand}, booktitle={2021 IEEE International Conference on Automatic Face and Gesture Recognition (FG)}, title={Detecting Human-to-Human-or-Object (H2O) Interactions with DIABOLO}, year={2021}, volume={}, number={}, pages={}, doi={}}

ImagesH²O 数据集由两部分图像组成：其一为取自V-COCO数据集（V-COCO）的10301张图像，其二为额外增补的3635张以人物交互为主要内容的图像。

额外图像的采集来源包括：MS-COCO数据集（MS-COCO）、Human-Interaction-Images数据集（Human-Interaction-Images）、BIT-Interaction数据集（BIT-Interaction）、SBU-Kinect-Interaction数据集（SBU-Kinect-Interaction）、TV-Human-interaction数据集（TV-Human-interaction）、Pascal-Voc数据集（Pascal-Voc）、ShakeFive2数据集（ShakeFive2）、UT-Interaction数据集（UT-Interaction）以及网络公开图像。

标注规范与分类体系
所有H²O图像均采用全新的动词分类体系进行标注，覆盖人与物体、人与人之间的交互行为。该分类体系共包含51个动词，划分为五大类别：
1. 描述主体通用姿态的动词
2. 描述主体移动方式的动词
3. 用于物体交互的动词
4. 描述人与人交互的动词
5. 涉及力量或暴力的交互动词

引用信息
A. Orcesi、R. Audigier、F. Poka Toukam与B. Luvison，《Detecting Human-to-Human-or-Object (H2O) Interactions with DIABOLO》，2021 IEEE国际自动人脸与手势识别会议（FG2021），https://arxiv.org/abs/2201.02396

对应的BibTeX引用格式如下：
@INPROCEEDINGS{h2o_dataset_2021,
author={Orcesi, Astrid and Audigier, Romaric and Poka Toukam, Fritz and Luvison, Bertrand},
booktitle={2021 IEEE International Conference on Automatic Face and Gesture Recognition (FG)},
title={Detecting Human-to-Human-or-Object (H2O) Interactions with DIABOLO},
year={2021},
volume={},
number={},
pages={},
doi={}
}

搜集汇总

背景与挑战

背景概述

H²O行人交互检测数据集是一个专注于检测人与人或人与物体互动的数据集，包含13,936张图片，标注了51个动词分类。该数据集来源于多个公开数据集和网络图片，适用于目标检测任务，许可类型为CC BY-NC。