search

happynear / FaceVerification

An Experimental Implementation of Face Verification, 96.8% on LFW.
447 stars 234 forks source link

about face align #50

Open ZHAIXINGZHAIYUE opened 7 years ago

ZHAIXINGZHAIYUE commented 7 years ago

大神您好。aligned_lfw.zip 中矫正好的图片您是用什么方法得到的呢？ 我在重现ydwen的实验时，我用自己矫正的lfw数据集做测试，accuracy只有0.981,而用您矫正的LFW做测试时accuracy有0.988。 我使用的是mtcnn做的人脸检测和关键点检测。

happynear commented 7 years ago

this code https://github.com/happynear/FaceVerification/blob/master/dataset/general_align.m .

ZHAIXINGZHAIYUE commented 7 years ago

your code works well, thank you.The method to align face : wuxiang or wenyandong,which is better? Could you give me your advance? Thanks again.

happynear commented 7 years ago

Wu's alignment is different with Wen's alignment. Wu's aligned image is of size 144144, while Wen's is 96112.

happynear commented 7 years ago

If you are using Wen's model, you must use his alignment method.

iamZe commented 7 years ago

@happynear @ZHAIXINGZHAIYUE 请问align LFW images的时候，使用general_align.m在LFW的lfw2 folder里（也就是raw images）还是应用在lfw_funneled或者lfw-deepfunneled上呢？应该会有的人脸检测不到吧

happynear commented 7 years ago

@iamZe ，我记得是有样本检测不出来的，我当时是随便画了个区域，直接使用第三个和第四个网络来得到keypoints。我记得标出来的点还是挺准的。