Open ZiaLee opened 5 years ago
你把一张图的多个groud truth加起来,取平均,用平均值大于0.5的作为正样本,平均值为0的作为负样本,平均值在0到0.5之间的忽略掉。这就是RCF论文在第3.2节说的处理方法。你可以用这个工程里的Caffe,处理后的数据已经在这个工程下面公布了,你再仔细看下,https://github.com/yun-liu/rcf,training RCF的第3节,链接http://mftp.mmcheng.net/liuyun/rcf/data/HED-BSDS.tar.gz
你好问一下关于ODS F-Measure的计算在代码的哪个文件里,因为不熟悉caffe没有找到,谢谢
我代码里没有这个, 你可以去他们的BSDS500数据集主页上下载 https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html
感谢你提供的数据集! @yun-liu
你把一张图的多个groud truth加起来,取平均,用平均值大于0.5的作为正样本,平均值为0的作为负样本,平均值在0到0.5之间的忽略掉。这就是RCF论文在第3.2节说的处理方法。你可以用这个工程里的Caffe,处理后的数据已经在这个工程下面公布了,你再仔细看下,https://github.com/yun-liu/rcf,training RCF的第3节,链接http://mftp.mmcheng.net/liuyun/rcf/data/HED-BSDS.tar.gz
您好,我下载您处理后的数据,但是test对应的GT没有给出?方便再公开些嘛??谢谢。
test的数据可以从那些数据原有的官网上下载的。 BSDS500 (官网): https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html NYUD (HED的数据): http://vcl.ucsd.edu/hed/nyu/ Multicue and PASCAL (HED的数据): http://vcl.ucsd.edu/datasets/
test的数据可以从那些数据原有的官网上下载的。 BSDS500 (官网): https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html NYUD (HED的数据): http://vcl.ucsd.edu/hed/nyu/ Multicue and PASCAL (HED的数据): http://vcl.ucsd.edu/datasets/
Cannot get access to http://vcl.ucsd.edu/datasets/ ? Any advice please?
I cannot access http://vcl.ucsd.edu/datasets/ either. Please follow the instructions on https://github.com/yun-liu/RCF to download data. For the Multicue dataset, you can refer to this website: https://serre-lab.clps.brown.edu/resource/multicue/.
I cannot access http://vcl.ucsd.edu/datasets/ either. Please follow the instructions on https://github.com/yun-liu/RCF to download data. For the Multicue dataset, you can refer to this website: https://serre-lab.clps.brown.edu/resource/multicue/.
Sorry for disturbance again. Did you kept the groundtruth file (in .mat) of multicue dataset offline?
If it is convenience for you, could you please offer me them, in any platform you wish?
@rrryan2016 I don't have the .mat version of the groundtruth.
@rrryan2016 I don't have the .mat version of the groundtruth.
I gerenated it by myself. Thanks for everything!
You are welcome!
在处理数据集将groundtruth的.mat文件转换为.jpg时,发现groudtruth中有多个Boundaries,问一下这个应该使用哪一个或者使用并集?如果可以的话能不能发一下你处理后的数据集,还有一些其他问题能不能发一下你的邮箱?谢谢 @yun-liu