search

jwyang / face-alignment

Face alignment in 3000 FPS
322 stars 190 forks source link

A problem about "draw.m"? #7

Open weiyt2014 opened 9 years ago

weiyt2014 commented 9 years ago

In globalpredicition.m,"drawshapes(Te_Data{k}.img_gray, [Te_Data{k}.shape_gt Te_Data{k}.intermediate_shapes{1}(:,:, s) shape_newstage]);" Te_Data{k}.shape_gt ,if it not exists,how to get the landmarks points?

jwyang commented 9 years ago

You do not have to get the groundtruth landmarks. If you have no groundtruth landmarks, you can remove Te_Data{k}.shape_gt from that line.

在 2014-12-08 11:02:01，"weiyt2014" notifications@github.com 写道：

In globalpredicition.m,"drawshapes(Te_Data{k}.img_gray, [Te_Data{k}.shape_gt Te_Data{k}.intermediate_shapes{1}(:,:, s) shape_newstage]);" Te_Data{k}.shape_gt ,if it not exists,how to get the landmarks points?

— Reply to this email directly or view it on GitHub.

weiyt2014 commented 9 years ago

I have removed it ,but the effect is bad.I don't know the problem is where?mainly predicition landmarks are errors and are very different from ideal results.

jwyang commented 9 years ago

yes, if it is true, you may need to re-train the model. By th way, what is your training error?

weiyt2014 commented 9 years ago

training error is often between 0.1 and 0.2,but some also 0.5,0.6!

weiyt2014 commented 9 years ago

Have you the training model? If possible ,could you send me a copy one.

ÔÚ 2014-12-08 22:35:45£¬"jwyang" notifications@github.com дµÀ£º

yes, if it is true, you may need to re-train the model. By th way, what is your training error?

¡ª Reply to this email directly or view it on GitHub.