无法复现论文结果

[Vincent]

作者你好，我运行了你的代码。
按你论文中提到的，用了MCG的superpixels作为输入，默认参数按照你的代码没有修改，并将结果保存为了mat进行eval。
我跑出来的结果如下，无法复现论文中的结果
SC=0.43，GT=0.49，RI=0.72，VI=2.23
请问是我参数还是哪些操作出了问题吗？谢谢

[zmbhou]

Hi, I am not sure what kind of MCG superpixels you are using. I will try to upload the baseline superpixels used in my paper. It is noted that the results reported are based on the best segmentation selection with multiple settings. The results using fixed SPN are displayed in Table. 4, which are lower than the OIS results reported in Table. 1.

[Vincent]

Okey, i will try.
by the way, are you using the BSDS500 benchmark for evaluate? like here https://github.com/BIDS/BSDS500.
i want to know how to compute the metric like GCE or BDE, can you tell me the code about the evaluate.

[Vincent]

thank you!

[zmbhou]

Please refer to the source code of SAS or MLSS, you can find the codes for computing GCE or BDE.

[Vincent]

okey, thank for your reply