Xin Pan, Ce Zhang, Jun Xu, Jian Zhao, 2021, Simplified Object-Based Deep Neural Network for Very High Resolution Remote Sensing Image Classification, ISPRS Journal of Photogrammetry and Remote Sensing, 181:218-237.
SO-DNNAlgorithmScripts contains Scripts of all SO-DNN algorithms.
PixelSampleSelecterTools can generate pixle-based samples for algorithm operation.
For the python script of the algorithm, TensorFlow and SCIKit-image are required.
In all scripts, running A1~A8 in sequence can perform the method described in the paper, where not running A6 and A7 corresponds to the output of the first iteration of SO-DNN, and running around of A6 and A7 corresponds to the output of the second iteration, the more running rounds, the more iterations.