Dhruvvv26/CrowdHuman
收藏资源简介：
--- license: cc-by-nc-4.0 task_categories: - object-detection language: - en pretty_name: CrowdHuman size_categories: - 10K<n<100K --- # CrowdHuman: A Benchmark for Detecting Human in a Crowd - 🏠 homepage: https://www.crowdhuman.org/ - 📄 paper: https://arxiv.org/pdf/1805.00123 CrowdHuman is a benchmark dataset to better evaluate detectors in crowd scenarios. The CrowdHuman dataset is large, rich-annotated and contains high diversity. CrowdHuman contains 15000, 4370 and 5000 images for training, validation, and testing, respectively. There are a total of 470K human instances from train and validation subsets and 23 persons per image, with various kinds of occlusions in the dataset. Each human instance is annotated with a head bounding-box, human visible-region bounding-box and human full-body bounding-box. We hope our dataset will serve as a solid baseline and help promote future research in human detection tasks.  *Volume, density and diversity of different human detection datasets. For fair comparison, we only show the statistics of training subset.* ## 🔍 Samples |||| |:--:|:--:|:--:| |||| |||| ## 📁 Files - `CrowdHuman_train01.zip` - `CrowdHuman_train02.zip` - `CrowdHuman_train03.zip` - `CrowdHuman_val.zip` - `CrowdHuman_test.zip` - `annotation_train.odgt` - `annotation_val.odgt` ## 🖨 Data Format We support `annotation_train.odgt` and `annotation_val.odgt` which contains the annotations of our dataset. ### What is odgt? `odgt` is a file format that each line of it is a JSON, this JSON contains the whole annotations for the relative image. We prefer using this format since it is reader-friendly. ### Annotation format ```json JSON{ "ID" : image_filename, "gtboxes" : [gtbox], } gtbox{ "tag" : "person" or "mask", "vbox": [x, y, w, h], "fbox": [x, y, w, h], "hbox": [x, y, w, h], "extra" : extra, "head_attr" : head_attr, } extra{ "ignore": 0 or 1, "box_id": int, "occ": int, } head_attr{ "ignore": 0 or 1, "unsure": int, "occ": int, } ``` - `Keys` in `extra` and `head_attr` are **optional**, it means some of them may not exist - `extra/head_attr` contains attributes for `person/head` - `tag` is `mask` means that this box is `crowd/reflection/something like person/...` and need to be `ignore`(the `ignore` in `extra` is `1`) - `vbox, fbox, hbox` means `visible box, full box, head box` respectively ## ⚠️ Terms of use: by downloading the image data you agree to the following terms: 1. You will use the data only for non-commercial research and educational purposes. 2. You will NOT distribute the above images. 3. Megvii Technology makes no representations or warranties regarding the data, including but not limited to warranties of non-infringement or fitness for a particular purpose. 4. You accept full responsibility for your use of the data and shall defend and indemnify Megvii Technology, including its employees, officers and agents, against any and all claims arising from your use of the data, including but not limited to your use of any copies of copyrighted images that you may create from the data. ## 🏆 Related Challenge - [Detection In the Wild Challenge Workshop 2019](https://www.objects365.org/workshop2019.html) ## 📚 Citaiton Please cite the following paper if you use our dataset. ``` @article{shao2018crowdhuman, title={CrowdHuman: A Benchmark for Detecting Human in a Crowd}, author={Shao, Shuai and Zhao, Zijian and Li, Boxun and Xiao, Tete and Yu, Gang and Zhang, Xiangyu and Sun, Jian}, journal={arXiv preprint arXiv:1805.00123}, year={2018} } ``` ## 👥 People - [Shuai Shao*](https://www.sshao.com/) - [Zijian Zhao*](https://scholar.google.com/citations?user=9Iv3NoIAAAAJ) - Boxun Li - [Tete Xiao](https://tetexiao.com/) - [Gang Yu](https://www.skicyyu.org/) - [Xiangyu Zhang](https://scholar.google.com/citations?user=yuB-cfoAAAAJ) - [Jian Sun](https://scholar.google.com/citations?user=ALVSZAYAAAAJ)
--- license: 知识共享署名-非商业性使用4.0国际许可协议（CC BY-NC 4.0） task_categories: - 目标检测（object-detection） language: - 英语（en） pretty_name: CrowdHuman size_categories: - 10000 < 样本数 < 100000 --- # CrowdHuman：面向密集人群场景的人体检测基准数据集 - 🏠 项目主页：https://www.crowdhuman.org/ - 📄 论文链接：https://arxiv.org/pdf/1805.00123 CrowdHuman是一款专为密集人群场景下的检测器性能评估打造的基准数据集。该数据集规模庞大、标注详尽且具备高度多样性，其中训练集、验证集与测试集分别包含15000张、4370张与5000张图像。训练集与验证集总计涵盖47万个标注人体实例，单张图像平均标注人体数量约为23个，且数据覆盖多种遮挡场景。每个人体实例均标注有头部边界框（bounding-box）、人体可见区域边界框与人体全身边界框。本数据集旨在作为可靠的基准基线，助力人体检测领域的未来研究发展。  *不同人体检测数据集的样本体量、密度与多样性对比。为保证公平比较，此处仅展示训练集的统计数据。* ## 🔍 样本示例 |||| |:--:|:--:|:--:| |||| |||| ## 📁 数据集文件 - `CrowdHuman_train01.zip`：训练集分卷压缩包 - `CrowdHuman_train02.zip`：训练集分卷压缩包 - `CrowdHuman_train03.zip`：训练集分卷压缩包 - `CrowdHuman_val.zip`：验证集压缩包 - `CrowdHuman_test.zip`：测试集压缩包 - `annotation_train.odgt`：训练集标注文件 - `annotation_val.odgt`：验证集标注文件 ## 🖨 数据格式 本数据集提供`annotation_train.odgt`与`annotation_val.odgt`两类标注文件，涵盖全量数据集的标注信息。 ### 何为odgt格式？ `odgt`是一种每行均为单个JSON对象的文件格式，每个JSON对象包含对应图像的完整标注信息。我们选用该格式的原因在于其具备良好的可读性。 ### 标注格式 json JSON{ "ID" : image_filename, "gtboxes" : [gtbox], } gtbox{ "tag" : "person" or "mask", "vbox": [x, y, w, h], "fbox": [x, y, w, h], "hbox": [x, y, w, h], "extra" : extra, "head_attr" : head_attr, } extra{ "ignore": 0 or 1, "box_id": int, "occ": int, } head_attr{ "ignore": 0 or 1, "unsure": int, "occ": int, } - `extra`与`head_attr`中的键为可选字段，即部分字段可能不会出现 - `extra`与`head_attr`分别存储人体与头部的属性信息 - 当`tag`取值为`mask`时，表示该边界框对应人群、反光或类人体等非标准人体目标，需予以忽略（对应`extra`中的`ignore`字段取值为1） - `vbox`、`fbox`与`hbox`分别代表可见区域框、全身框与头部框 ## ⚠️ 使用条款 下载本图像数据集即代表您同意以下使用条款： 1. 您仅可将数据集用于非商业性研究与教育用途。 2. 您不得分发上述图像数据集。 3. 旷视科技（Megvii Technology）不对本数据集做出任何明示或默示担保，包括但不限于不侵权担保或特定用途适用性担保。 4. 您需对本数据集的使用承担全部责任，并应就因您使用数据集而引发的任何及所有索赔，为旷视科技（Megvii Technology）及其雇员、高管与代理人提供辩护与赔偿，包括但不限于您因使用本数据集生成的受版权保护的图像副本所引发的索赔。 ## 🏆 相关赛事 - [2019年野外检测挑战赛研讨会（Detection In the Wild Challenge Workshop 2019）](https://www.objects365.org/workshop2019.html) ## 📚 引用声明 若您使用本数据集，请引用以下论文： @article{shao2018crowdhuman, title={CrowdHuman: A Benchmark for Detecting Human in a Crowd}, author={Shao, Shuai and Zhao, Zijian and Li, Boxun and Xiao, Tete and Yu, Gang and Zhang, Xiangyu and Sun, Jian}, journal={arXiv preprint arXiv:1805.00123}, year={2018} } ## 👥 作者团队 - [邵帅*（Shuai Shao）](https://www.sshao.com/) - [赵子健*（Zijian Zhao）](https://scholar.google.com/citations?user=9Iv3NoIAAAAJ) - 李博坤（Boxun Li） - [肖特特（Tete Xiao）](https://tetexiao.com/) - [余钢（Gang Yu）](https://www.skicyyu.org/) - [张祥雨（Xiangyu Zhang）](https://scholar.google.com/citations?user=yuB-cfoAAAAJ) - [孙剑（Jian Sun）](https://scholar.google.com/citations?user=ALVSZAYAAAAJ)
