This is a dataset of touching cell images used for "A robust concave point detector based on k-variance of k-sum of chain code", which can be used to evaluate concave points detection methods.
The image dataset consists of 24 different original 300×300 binary images, which were collected from the CellaVision blog (http://blog.cellavision.com/) and segmented by our earlier work. Each image contains a single touching-cell clump.
All the concave points of each image were manually labeled by domain experts and stored in a .mat file. You can read the .mat files using MATLAB load statement as follows:
C = load('001.mat','Corner');
C is a m*2 matrix, where m is the total number of all the concave points in "001.bmp". Each row vector of C is the position (x,y) of each concave point.
If you use the dataset in your paper, please acknowledge the CellaVision blog <http://blog.cellavision.com/> (http://blog.cellavision.com/) and cite the paper:
@article{Zheng2018,
title={Fast and Robust Segmentation of White Blood Cell Images by Self-supervised Learning},
author={Xin Zheng and Yong Wang and Guoyou Wang and Jianguo Liu},
journal={Micron},
volume={107},
pages={55--71},
year={2018},
publisher={Elsevier}
doi={https://doi.org/10.1016/j.micron.2018.01.010},
url={https://www.sciencedirect.com/science/article/pii/S0968432817303037}
}