Harvard-Edge/Wake-Vision
收藏资源简介:
--- license: cc-by-4.0 size_categories: - 1M<n<10M task_categories: - image-classification pretty_name: Wake Vision dataset_info: features: - name: age_unknown dtype: class_label: names: '0': 'No' '1': 'Yes' - name: body_part dtype: class_label: names: '0': 'No' '1': 'Yes' - name: bright dtype: class_label: names: '0': 'No' '1': 'Yes' - name: dark dtype: class_label: names: '0': 'No' '1': 'Yes' - name: depiction dtype: class_label: names: '0': 'No' '1': 'Yes' - name: far dtype: class_label: names: '0': 'No' '1': 'Yes' - name: filename dtype: string - name: gender_unknown dtype: class_label: names: '0': 'No' '1': 'Yes' - name: image dtype: image - name: medium_distance dtype: class_label: names: '0': 'No' '1': 'Yes' - name: middle_age dtype: class_label: names: '0': 'No' '1': 'Yes' - name: near dtype: class_label: names: '0': 'No' '1': 'Yes' - name: non-person_depiction dtype: class_label: names: '0': 'No' '1': 'Yes' - name: non-person_non-depiction dtype: class_label: names: '0': 'No' '1': 'Yes' - name: normal_lighting dtype: class_label: names: '0': 'No' '1': 'Yes' - name: older dtype: class_label: names: '0': 'No' '1': 'Yes' - name: person dtype: class_label: names: '0': 'No' '1': 'Yes' - name: person_depiction dtype: class_label: names: '0': 'No' '1': 'Yes' - name: predominantly_female dtype: class_label: names: '0': 'No' '1': 'Yes' - name: predominantly_male dtype: class_label: names: '0': 'No' '1': 'Yes' - name: young dtype: class_label: names: '0': 'No' '1': 'Yes' splits: - name: train_quality num_bytes: 344101633249.375 num_examples: 1196221 - name: validation num_bytes: 5296900547.25 num_examples: 18582 - name: test num_bytes: 15830126041.625 num_examples: 55763 download_size: 26134101710 dataset_size: 21127026588.875 configs: - config_name: default data_files: - split: train_quality path: data/train_quality* - split: test path: data/test-* - split: validation path: data/validation-* --- # Dataset Card for Wake Vision ### Dataset Description <!-- Provide a longer summary of what this dataset is. --> "Wake Vision" is a large, high-quality dataset featuring over 6 million images, significantly exceeding the scale and diversity of current tinyML datasets (100x). This dataset includes images with annotations of whether each image contains a person. Additionally, it incorporates a comprehensive fine-grained benchmark to assess fairness and robustness, covering perceived gender, perceived age, subject distance, lighting conditions, and depictions. The Wake Vision labels are derived from Open Image's annotations which are licensed by Google LLC under CC BY 4.0 license. The images are listed as having a CC BY 2.0 license. Note from Open Images: "while we tried to identify images that are licensed under a Creative Commons Attribution license, we make no representations or warranties regarding the license status of each image and you should verify the license for each image yourself." - **License:** [CC-BY 4.0] ### Dataset Sources <!-- Provide the basic links for the dataset. --> - **Train (Large) Dataset:** https://huggingface.co/datasets/Harvard-Edge/Wake-Vision-Train-Large - **Website:** https://wakevision.ai/ - **Repository:** https://github.com/colbybanbury/Wake_Vision_Quickstart - **Paper:** https://arxiv.org/abs/2405.00892 ## Citation <!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. --> **BibTeX:** ```bibtex @misc{banbury2024wake, title={Wake Vision: A Large-scale, Diverse Dataset and Benchmark Suite for TinyML Person Detection}, author={Colby Banbury and Emil Njor and Matthew Stewart and Pete Warden and Manjunath Kudlur and Nat Jeffries and Xenofon Fafoutis and Vijay Janapa Reddi}, year={2024}, eprint={2405.00892}, archivePrefix={arXiv}, primaryClass={cs.CV} } ``` ## Dataset Card Contact For questions about the dataset or image takedown requests, contact wakevision@edgeaifoundation.org
license: CC BY 4.0 样本规模类别: - 100万 < 样本数量 < 1000万 任务类别: - 图像分类 数据集名称:Wake Vision dataset_info: 特征项: - 名称:age_unknown,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:body_part,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:bright,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:dark,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:depiction,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:far,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:filename,数据类型:字符串 - 名称:gender_unknown,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:image,数据类型:图像 - 名称:medium_distance,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:middle_age,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:near,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:non-person_depiction,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:non-person_non-depiction,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:normal_lighting,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:older,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:person,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:person_depiction,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:predominantly_female,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:predominantly_male,数据类型:分类标签,类别映射:'0': '否', '1': '是' - 名称:young,数据类型:分类标签,类别映射:'0': '否', '1': '是' 数据子集: - 名称:train_quality,字节数:344101633249.375,样本数量:1196221 - 名称:validation,字节数:5296900547.25,样本数量:18582 - 名称:test,字节数:15830126041.625,样本数量:55763 下载总大小:26134101710 数据集总存储大小:21127026588.875 配置项: - 配置名称:default,数据文件: - 子集train_quality:路径为data/train_quality* - 子集test:路径为data/test-* - 子集validation:路径为data/validation-* --- # Wake Vision 数据集卡片 ## 数据集描述 <!-- 请提供该数据集的详细摘要 --> Wake Vision是一款大规模高质量数据集,涵盖超过600万张图像,其规模与多样性显著超越当前微型机器学习(tinyML)数据集,达到后者的百倍之多。该数据集为每张图像标注了是否包含人物,同时还配备了一套全方位细粒度基准测试套件,用于评估模型的公平性与鲁棒性,覆盖感知性别、感知年龄、拍摄距离、光照条件以及图像表征等多个维度。 Wake Vision的标签源自Open Image数据集的标注,该标注由Google LLC以CC BY 4.0许可协议授权。数据集内的图像标注为遵循CC BY 2.0许可协议。Open Image的官方声明如下:「尽管我们已尽力识别采用知识共享署名许可协议的图像,但我们不对每张图像的许可状态作出任何陈述或保证,您需自行验证每张图像的许可协议。」 - **许可协议:** [CC-BY 4.0] ## 数据集来源 <!-- 请提供该数据集的基础链接 --> - **大型训练子集:** https://huggingface.co/datasets/Harvard-Edge/Wake-Vision-Train-Large - **官方网站:** https://wakevision.ai/ - **代码仓库:** https://github.com/colbybanbury/Wake_Vision_Quickstart - **相关论文:** https://arxiv.org/abs/2405.00892 ## 引用格式 <!-- 若该数据集有对应的论文或博客文章,请在此处添加其APA和BibTeX引用信息 --> **BibTeX格式:** bibtex @misc{banbury2024wake, title={Wake Vision: A Large-scale, Diverse Dataset and Benchmark Suite for TinyML Person Detection}, author={Colby Banbury and Emil Njor and Matthew Stewart and Pete Warden and Manjunath Kudlur and Nat Jeffries and Xenofon Fafoutis and Vijay Janapa Reddi}, year={2024}, eprint={2405.00892}, archivePrefix={arXiv}, primaryClass={cs.CV} } ## 数据集卡片联系方式 若对该数据集有疑问或需申请移除图像,请联系 wakevision@edgeaifoundation.org
数据集概述
- 名称: Wake Vision
- 任务类别: 图像分类
- 数据集大小: 1M<n<10M
- 许可: CC-BY 4.0
数据集特征
- age_unknown: 类别标签,0: No, 1: Yes
- body_part: 类别标签,0: No, 1: Yes
- bright: 类别标签,0: No, 1: Yes
- dark: 类别标签,0: No, 1: Yes
- depiction: 类别标签,0: No, 1: Yes
- far: 类别标签,0: No, 1: Yes
- filename: 字符串类型
- gender_unknown: 类别标签,0: No, 1: Yes
- image: 图像类型
- medium_distance: 类别标签,0: No, 1: Yes
- middle_age: 类别标签,0: No, 1: Yes
- near: 类别标签,0: No, 1: Yes
- non-person_depiction: 类别标签,0: No, 1: Yes
- non-person_non-depiction: 类别标签,0: No, 1: Yes
- normal_lighting: 类别标签,0: No, 1: Yes
- older: 类别标签,0: No, 1: Yes
- person: 类别标签,0: No, 1: Yes
- person_depiction: 类别标签,0: No, 1: Yes
- predominantly_female: 类别标签,0: No, 1: Yes
- predominantly_male: 类别标签,0: No, 1: Yes
- young: 类别标签,0: No, 1: Yes
数据集分割
- validation: 18582个样本,大小为5296900547.25字节
- test: 55763个样本,大小为15830126041.625字节
数据集大小
- 下载大小: 26134101710字节
- 数据集大小: 21127026588.875字节




