The meaning of measures in parameter optmization file

davidstutz / superpixel-benchmark

An extensive evaluation and comparison of 28 state-of-the-art superpixel algorithms on 5 datasets.

398 stars 109 forks source link

The meaning of measures in parameter optmization file #5

Closed kaiwensophie closed 4 years ago

kaiwensophie commented 5 years ago

Hi, I'm working on the parameter optimization of superpixels. After run the eval_parameter_optimization_cli command, the metrics for different parameters, and the optimal parameters are listed in parameter_optimazation.csv file. In this file, I wonder what does score, co_score, and sp_average mean. And what is the difference between best and co_best? Thank you.

davidstutz commented 4 years ago

The score is the average of UE and Boundary Recall, that's what we optimized for (see here: https://github.com/davidstutz/superpixel-benchmark/blob/master/lib_eval/parameter_optimization_tool.cpp#L408). Similary you can find the other metrics in the same file, e.g. here for co_score: https://github.com/davidstutz/superpixel-benchmark/blob/master/lib_eval/parameter_optimization_tool.cpp#L442.