This benchmark database is used for small target detection in the following paper:
Jing Lou, Wei Zhu, Huan Wang, Mingwu Ren, "Small Target Detection Combining Regional Stability and Saliency in a Color Image," Multimedia Tools and Applications, vol. 76, no. 13, pp. 14781-14798, 2017. doi:10.1007/s11042-016-4025-7
Project page: http://www.loujing.com/rss-small-target/
You can directly download the zipped file of the database: Benchmark.zip.
Copyright (C) 2016 Jing Lou (楼竞)
The usage of this data and code is restricted for non-profit research usage only and using of the code is at the user's risk.
This benchmark database contains three datasets (totally 1,093 color images). Each dataset folder includes the following subfolders:
Imageinput color image
RSStwo example models for reference
This benchmark database provides the pixel-wise ground truth for each input image. In
GT, each MAT file contains four properties of the ground truth target: Area, Boundary, Bounding Box, and Pixel List.
You can use the MATLAB script Evaluation.m to evaluate your detection results. Please read the help text in the script for more details.