Person-Specific-FAS icon indicating copy to clipboard operation
Person-Specific-FAS copied to clipboard

abolutely can't run your code following your instruction

Open Jade999 opened this issue 7 years ago • 17 comments

I think your instruction about how to run the code may be lack one important step,which may be plugged after "train models", because after trained models, when I go into Test_Models and run "runall_ps.m" or "runall_gen.m", it comes up one prompt as follows:

"Unable to read file '../Develop_Models/GenericPerf_devel_withSubNum_HOG_PCA_Enroll.mat'. No such file or directory.
  Error in test_GenericModel_withSubNum_specific_quality (line 232)
load(strcat('../Develop_Models/GenericPerf_devel_withSubNum_', name{Feat_Type}, '_', method, '_', suffix, '.mat'));"

Then I think, I may shoud run the Develop_Models/runall_ps.m or runall_gen.m first, but there is another wrong prompt

" Error using  + Matrix dimensions must agree. 
 Error in devel_FASModels_specific_quality (line 249)   scores = scores_specific + scores_generic;"

It was caused at line 247 in devel_FASModels_specific_quality.m, of which the parameter "GenericModel" is a 15*1 cell, not a model struct.

At last,may I offer you a proposal in good faith that may you purge the code before release, so many commented code may perplex others who want follow your code. Thank you a lot.

Jade999 avatar Mar 31 '17 07:03 Jade999

Hi, thanks for pointing out this bug and your suggestions. I will debug the code and clean the code in the following days.

jwyang avatar Mar 31 '17 15:03 jwyang

你好,请问一下,你这个最后测试效果怎么样。最好有一个明确一点的说明文档。

okmmsky888 avatar Apr 07 '17 08:04 okmmsky888

@caoyushe Hello, can you give me a good model of training? I trained to GenericModel_withSubNum_MsLBP_PLS_PCA this model, the memory is not enough,thank you

okmmsky888 avatar Apr 08 '17 14:04 okmmsky888

@okmmsky888 sorry, I don't have any trained model. Due to an unexpected bug, the training process was interrupted...

Jade999 avatar Apr 10 '17 03:04 Jade999

@jwyang hi请问下,你的这个代码,我目前在自己的本机上进行训练,我内存是12G,但是训练GenericModel_withSubNum_MsLBP_PLS_PCA这个的时候,内存还是不够用了。 在代码中,我看到很多内容都被注释了。也没有进行明确的说明。你有已经训练好的模型吗?如果有的话,能否提供给我。

okmmsky888 avatar Apr 12 '17 08:04 okmmsky888

@caoyushe OK

okmmsky888 avatar Apr 12 '17 08:04 okmmsky888

@okmmsky888

Hi, 这个project是三年前在nlpr做的,之前的服务器上的代码和模型估计已经清空了,我只拷贝了代码和提取的特征,没有保存pre-trained模型。如果要训练完整的模型的话,可能还是需要在服务器上跑。PLS—_PCA setting比较耗费内存,你可以试试center-shift (CS) setting下训练,耗费的内存比较小。

jwyang avatar Apr 13 '17 23:04 jwyang

@jwyang 那之前的这套代码,你们测试效果怎么样,准确率能够达到多少?还有你们提取的特征都是哪些基本的特征

okmmsky888 avatar Apr 14 '17 03:04 okmmsky888

@okmmsky888 ,这个代码就是tifs文章的效果了,没动过。提取的特征你可以从google drive上下载,里面包含了两种特征:mslbp和hog

jwyang avatar Apr 14 '17 03:04 jwyang

@jwyang OK,谢谢

okmmsky888 avatar Apr 14 '17 03:04 okmmsky888

@jwyang 你好,我在问一下,针对CASIA数据集,matlab中怎么没有Estimate_Transformation模块? replay-attack中的Estimate_Transformation模块,是否可以直接用在CASIA数据集上,进行Transformation

okmmsky888 avatar Apr 14 '17 04:04 okmmsky888

@okmmsky888 我当时也找了半天没找到,当时从服务器传到网盘时可能落了这个文件夹了。。。这个好像不能直接share,因为两个数据库的spoofing type有区别,变量都不太一样,可能需要改一下才行。我看看有没有空改一改,不过这个project的code现在基本都忘记了,需要点时间。你可以先在Replay-Attack上面跑一下。

jwyang avatar Apr 14 '17 04:04 jwyang

@jwyang OK,还是希望你抽出时间,补全一下CASIA数据模块的代码,使其能够完整分享

okmmsky888 avatar Apr 14 '17 06:04 okmmsky888

恩,谢谢你的建议

jwyang avatar Apr 14 '17 07:04 jwyang

@jwyang 你好,我在Test_Models 下运行runall_gen.m得到GenericPerf_test_withSubNum_MsLBP_CS_Enroll,GenericPerf_test_withSubNum_HOG_CS_Enroll 等 几个.mat文件,请问一下,这几个mat文件,是存放的测试结果吗?这里面存放了好多值,还有Develop_Models这个文件夹下面的的模型跟测试模型什么关系啊。感觉有点乱,能否解释一下,谢谢

okmmsky888 avatar Apr 19 '17 12:04 okmmsky888

不好意思,我好久没做反欺诈了

发自我的 iPhone

在 2018年10月8日,16:33,weihua04 [email protected] 写道:

您好,请教个问题哈~针对纸质照片和电子屏显示照片的伪造攻击,有没有速度快效果差不多的特征? 请问有没有训练好的分类器模型能否上传下或者发给我呢[email protected]?非常感谢

― You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

Jade999 avatar Oct 08 '18 08:10 Jade999

恩,谢谢你的建议

@jwyang 您好,请教个问题哈~针对纸质照片和电子屏显示照片的伪造攻击,有没有速度快效果差不多的特征? 请问有没有训练好的分类器模型能否上传下或者发给我呢[email protected]?非常感谢

weihua04 avatar Oct 08 '18 08:10 weihua04