mtcnn ROC question

Hi: I notice that you train the mtcnn and release model, can you tell me your roc? Compared with the original model, how about performance of your model?

Thanks

Dec 09 '16 06:12 kingvision

The recall of p-net and r-net are slightly lower(1%～%2, minimum detect face size=24) than the original model while having the same false positives, so I use lower threshold to keep the recall. The roc curve of o-net is almost the same as the original one.

Dec 09 '16 10:12 Seanlinx

how about the finnal ROC performance on FDDB please? Is it also the same as original one ?

Dec 16 '16 01:12 hiyijian

I have updated the final disROC. The performance will be better if joint landmark detection in training according to the paper.

Dec 16 '16 03:12 Seanlinx

Hi, @Seanlinx thresh setting: 0.6, 0.3, 0.1 is only for test phase, or during data preparation?

Dec 23 '16 06:12 mychina75

@mychina75 it's for testing. The thresholds I used for data preparation are a little lower just to cover more positive samples.

Dec 28 '16 12:12 Seanlinx

@Seanlinx ，hi，Can you tell me how to plot the PR curve for WIDER test?

Jun 12 '17 11:06 jianbaba