Obect Skeleton Detection Dataset with 1K+ Instances


This dataset, namely SK-LARGE, is used for evaluating the methods for skeleton extraction in natural images, which contents 1491 images selected from MS COCO [1], 746 for training and 745 for testing.

Each image in SK-LARGE has a corresponding skeleton map, which is extracted by using the approach introduced in [2]. Each groundtruth is represented by a .mat file, in which there are two fields:

pressue plot
Image showcase from SK-LARGE.

Skeleton Scale

The scale of a skeleton point is defined as its distance to the nearest edge point, which can be easily implemented by the Matlab function bwdist.

A Matlab script augmentation.m is attached for data augmentation (flip, rotate and crop the origin images as well as groundtruth maps).

How to Use

Download and do data augmentation through simple steps if you are on Linux with Matlab installed:

# download and extract data
tar -xvf sk1491.tar.gz
# data augmentation
matlab -nodisplay -r "run augmentation.m; exit"
skeval is the Code for skeleton-extraction performance evaluation.


If our method is helpful to your research, please kindly consider to cite:
  title        = {DeepSkeleton: Learning Multi-task Scale-associated Deep Side Outputs for Object Skeleton Extraction in Natural Images},
  author       = {Shen, Wei and Zhao, Kai and Jiang, Yuan and Wang, Yan and Bai, Xiang and Yuille, Alan},
  journal      = {IEEE Transactions on Image Processing},
  volume       = {26},
  number       = {11},
  pages        = {5298-5311},
  year         = {2017},
  publisher    = {IEEE},
  howpublished = "\url{}"
  title     = {Hi-{F}i: Hierarchical Feature Integration for Skeleton Detection},
  author    = {Kai Zhao and Wei Shen and Shanghua Gao and Dandan Li and Ming-Ming Cheng},
  booktitle = {Proceedings of the Twenty-Seventh International Joint Conference on
               Artificial Intelligence, {IJCAI-18}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  pages     = {1191--1197},
  year      = {2018},
  month     = {7},
  doi       = {10.24963/ijcai.2018/166},
  url       = {},
[1] X. Chen, H. Fang, T. Lin, R. Vedantam, S. Gupta, P. Dollar, and C. L. Zitnick. Microsoft COCO captions: Data collection and evaluation server. arXiv:1405.0312, 2015.

[2] Wei Shen, Xiang Bai, Xingwei Yang, Longin Jan Latecki. Skeleton Pruning as Trade-off between Skeleton Simplicity and Reconstruction Error. SCIENCE CHINA Information Sciences, 56(4): 1-14, 2013.