deep-head-pose icon indicating copy to clipboard operation
deep-head-pose copied to clipboard

different input have the same result

Open you-old opened this issue 6 years ago • 9 comments

i trained a model use hopenet, but when i test ,every input image have the same result ,why?

you-old avatar Jul 06 '19 06:07 you-old

hello! I have no the file about face detection, if you have it, please give it me, thank you!

yafeiliu666 avatar Jul 12 '19 07:07 yafeiliu666

Same problem here. Any updates?

howard5758 avatar Jul 18 '19 19:07 howard5758

Have you solved this problem? I have encountered the same problem.

ZhongWenY avatar Aug 06 '19 00:08 ZhongWenY

i trained a model use hopenet, but when i test ,every input image have the same result ,why?

Have you solved this problem? I have encountered the same problem.

ZhongWenY avatar Aug 06 '19 09:08 ZhongWenY

Continue training, i have this problem , but i continue training now it is ok ,you can try change your optimizer and continue training

guojilei avatar Sep 06 '19 07:09 guojilei

dear @guojilei, How to train with 300W_LP dataset from at first?

NguyenDangBinh avatar Nov 07 '19 05:11 NguyenDangBinh

Continue training, i have this problem , but i continue training now it is ok ,you can try change your optimizer and continue training

How many epochs did you train? and Did you change optimizer or used default of code

LamLuong avatar Feb 25 '21 11:02 LamLuong

Hi everyone, I resolved problem above, you should not use default alpha and learning rate value of code, should use this values in paper

LamLuong avatar Feb 26 '21 05:02 LamLuong

Hi everyone, I resolved problem above, you should not use default alpha and learning rate value of code, should use this values in paper

hello.Can you send me your final trained model? Thank you

x1ngzai avatar Jul 24 '21 12:07 x1ngzai