Results 45 comments of lfz

That't maybe because the lack of memory. reduce the cpu thread used for preprocessing

sorry, the official site closed this

I haven't tried it, but I think it would be a very very long time 2017-12-13 0:01 GMT-08:00 dongxuemin : > in other words,if I do not use cuda,does it...

Thank you for letting me know that, I think firstly any method would perform better with smaller slice thickness because of more information. but this difference is too big? so...

adapt_ckpt convert the parameters of detector to parameters of classifier main is the training program of classifier

I havent't seen this bug before. Maybe it's a pytorch bug? check the pytorch version? 2017-09-02 10:36 GMT+08:00 zhaifly : > python version: 2.7 > > BR. > -zhaifly >...

No idea, please provide more info? like the error log 2017-06-05 23:44 GMT+08:00 galactica147 : > Hi, > > Thanks for posting the code for us to reproduce the results!...

我不是很明白你的问题,你可以发中文 2018-01-30 11:43 GMT+08:00 yz275252130 : > I want to find the corresponding label for test.npy(DSB2017/training/detector/test.npy > ) > I see it in the paper.Because our detection module is designed...

这个198个病例的数据可以从DSB原网站的数据栏目下载,另外,这个198个不是luna的,是kaggle的数据,里边的标签是我们自己打的,并不准 在 2018年1月30日 下午5:02,yz275252130 写道: > 在该项目 对应的论文中 我看到如下描述:(论文地址https://arxiv.org/pdf/1711.08324.pdf) > 由于我们的检测模块设计为在训练期间忽略非常小的结节,所以LUNA16评估系统不是适合评估其性能。 > 我们评估了DSB验证集的性能。它包含198例病例的数据,总共有71个(小于6毫米的结节被排除)结节。 > 自由响应操作特性(FROC)曲线如图7a所示。每次扫描的1/8,1 / 4,1 / 2,1,2,4,8假阳性的平均回忆是0.8562。 > > 在项目代码中我找到了test.npy 文件(DSB2017 / training / detector / test.npy) >...

这些原始label数据全都在label文件夹那些表格文件里边 这个test npy 只适合这个crop和分辨率 在 2018年1月31日 上午10:40,yz275252130 写道: > 这个198个病例的数据 我已经下载了,希望你们能够把这198个病例对应的标签数据给我一份 谢谢了! > > — > You are receiving this because you commented. > Reply to this email directly, view...