search

sai-bi / FaceAlignment

Face Alignment by Explicit Shape Regression
MIT License
339 stars 205 forks source link

The purpose of the bounding box #27

Open Yopzhao opened 8 years ago

Yopzhao commented 8 years ago

Hello,I have trained on the facedatabase MUCT,and 1000 pictures are used of testing set,another 500 pictures are used as training set.The result of alignment is good,but when I use JAFFE as the testing set,the result is poor.I think that the reason is the bounding box,such as the follow picture,the bounding box as it is, image. But I don't kown where I went wrong?

mathCrazyy commented 7 years ago

我用的也是这个库训练..效果...呃呃呃呃疯了..你现在训练的怎么样了