Jianzhong He
Jianzhong He
@young532 我想你说的是关于消融实验部分的结果我们在文章中也说了，是在BSDS500的训练集上训练，在BSDS500的验证集上的验证结果。关于GT就是先将几个标注的做平均得到概率图，然后按照代码中提供的处理方式处理的。
你好，你说的这个问题在我的训练过程中没有出现过。我也训练过很多次了，包括多个不同的数据集，训练中收敛情况挺好的，但是 偶尔的确会出现不收敛，这种情况非常少。
Thanks for your attention, I'll upload the pretrained model to google drive as soon as possible.
As @xavysp said, since the training data is insufficient, the pretrained weight have small lr and new added weights has relatively large lr. You should note that the pretrained model...
according to your description > the left side of an edge are the same as the right side i think the problem should be the approach you preprocess the data,...
I've not calculation the mean value of BSDS500, the mean value takes from HED.
你可以通过提供的与训练模型以及测试代码，很容易生成最终的检测结果的
好的，我会尽快更新一下百度网盘链接
之前链接有效啊，下面是更新的链接 updated the model link Link: https://pan.baidu.com/s/1yhQ4GC9IcwC1NCqhGudxvg code: 1zan
If your edges get from segmentation, you need not care about the parameter yita. If you edges is probability in (0,1), the yita control the confidence that the edges of...