svnfs/depth-of-field

--- license: apache-2.0 annotations_creators: - Stavros Niafas sample_number: - 1200 class_number: - 2 image_size: - (200,300,3) source_dataset: - unsplash task_categories: - image-classification - image-segmentation dataset_info: - config_name: depth-of-field features: - name: image dtype: string - name: class dtype: class_label: names: 0: bokeh 1: no-bokeh - config_name: default features: - name: image dtype: image - name: label dtype: class_label: names: 0: '0' 1: '1' splits: - name: train num_bytes: 192150 num_examples: 1200 download_size: 38792692 dataset_size: 192150 --- ## Dataset Summary Depth-of-Field(DoF) dataset is comprised of 1200 annotated images, binary annotated with(0) and without(1) bokeh effect, shallow or deep depth of field. It is a forked data set from the [Unsplash 25K](https://github.com/unsplash/datasets) data set. ## Dataset Description - **Repository:** [https://github.com/sniafas/photography-style-analysis](https://github.com/sniafas/photography-style-analysis) - **Paper:** [More Information Needed](https://www.researchgate.net/publication/355917312_Photography_Style_Analysis_using_Machine_Learning) ### Citation Information ``` @article{sniafas2021, title={DoF: An image dataset for depth of field classification}, author={Niafas, Stavros}, doi= {10.13140/RG.2.2.29880.62722}, url= {https://www.researchgate.net/publication/364356051_DoF_depth_of_field_datase} year={2021} } ``` Note that each DoF dataset has its own citation. Please see the source to get the correct citation for each contained dataset.

原始信息汇总

数据集概述

基本信息

• 许可证: Apache-2.0
• 样本数量: 1200
• 类别数量: 2
• 图像尺寸: (200, 300, 3)
• 源数据集: Unsplash
• 任务类别:
  • 图像分类
  • 图像分割

数据集配置

• config_name: depth-of-field

  • 特征:
    • image: 数据类型为字符串
    • class: 数据类型为类别标签，包含两个类别：
      • 0: bokeh
      • 1: no-bokeh

• config_name: default

  • 特征:
    • image: 数据类型为图像
    • label: 数据类型为类别标签，包含两个类别：
      • 0: 0
      • 1: 1
  • 分割:
    • train: 包含1200个样本，占用192150字节

数据集大小

• 下载大小: 38792692字节
• 数据集大小: 192150字节

数据集描述

• 内容: 包含1200张带有深度场效果（bokeh）和无深度场效果（no-bokeh）的二进制标注图像。
• 来源: 该数据集是从Unsplash 25K数据集派生而来。