graviti
产品服务
解决方案
知识库
公开数据集
关于我们
AISegment.com - Matting Human Datasets
Aesthetics
|...
许可协议: CC-BY-SA 4.0

Overview

Context

Human segmentation, i.e. high resolution extraction of humans from images, is a fascinating application with many uses. However, the problem is significantly under-constrained, making it an active area of research for developing more advanced methods. This dataset, developed by AISegment aims to help by providing a solid quality dataset of images and masks.

Quoting from the dataset author's GitHub (translated via Google Translate):
> This dataset is currently the largest portrait matting dataset, containing 34,427 images and corresponding matting results.
The data set was marked by the high quality of Beijing Play Star Convergence Technology Co., Ltd., and the portrait soft segmentation model trained using this data set has been commercialized.
The original images in the dataset are from Flickr, Baidu, and Taobao. After face detection and area cropping, a half-length portrait of 600*800 was generated.
The clip_img directory is a half-length portrait image in the format jpg; the matting directory is the corresponding matting file (convenient to confirm the matting quality), the format is png, you should first extract the alpha map from the png image before training. For example, using opencv you can get an alpha map like this:

In_image = cv2.imread('png image file path', cv2.IMREAD_UNCHANGED)  
Alpha = in_image[:,:,3]

License

See the author's GitHub.

Content

This dataset comes in two parts:

  1. Full images
  2. The respective RGB "masks" or "cutouts" of those images

Acknowledgements

Thanks to the folks from SegmentAI for putting this dataset together.

数据概要
数据格式
image,
数据量
68.854K
文件大小
3.6GB
发布方
Laurent H.
| 数据量 68.854K | 大小 3.6GB
AISegment.com - Matting Human Datasets
Aesthetics
许可协议: CC-BY-SA 4.0

Overview

Context

Human segmentation, i.e. high resolution extraction of humans from images, is a fascinating application with many uses. However, the problem is significantly under-constrained, making it an active area of research for developing more advanced methods. This dataset, developed by AISegment aims to help by providing a solid quality dataset of images and masks.

Quoting from the dataset author's GitHub (translated via Google Translate):
> This dataset is currently the largest portrait matting dataset, containing 34,427 images and corresponding matting results.
The data set was marked by the high quality of Beijing Play Star Convergence Technology Co., Ltd., and the portrait soft segmentation model trained using this data set has been commercialized.
The original images in the dataset are from Flickr, Baidu, and Taobao. After face detection and area cropping, a half-length portrait of 600*800 was generated.
The clip_img directory is a half-length portrait image in the format jpg; the matting directory is the corresponding matting file (convenient to confirm the matting quality), the format is png, you should first extract the alpha map from the png image before training. For example, using opencv you can get an alpha map like this:

In_image = cv2.imread('png image file path', cv2.IMREAD_UNCHANGED)  
Alpha = in_image[:,:,3]

License

See the author's GitHub.

Content

This dataset comes in two parts:

  1. Full images
  2. The respective RGB "masks" or "cutouts" of those images

Acknowledgements

Thanks to the folks from SegmentAI for putting this dataset together.

0
立即开始构建AI
graviti
wechat-QR
长按保存识别二维码,关注Graviti公众号

Copyright@Graviti
沪ICP备19019574号
沪公网安备 31011002004865号