search

xipeng13 / recurrent-face-alignment

Xi Peng et al. "A Recurrent Encoder-Decoder Network for Sequential Face Alignment", ECCV 2016. (Oral)
61 stars 20 forks source link

"Error parsing text-format caffe.NetParameter: 48:12: Message type "caffe.LayerParameter" has no field named "bn_param" #7

Open nurvarli opened 6 years ago

nurvarli commented 6 years ago

Hi, When I run this code, I get some error like below:

Error parsing text-format caffe.NetParameter: 48:12: Message type "caffe.LayerParameter" has no field named "bn_param".

Could you help me?

ChiefGodMan commented 6 years ago

Have you solved this problem?

vkasojhaa commented 5 years ago

This is because you are not using the caffe version that support batch normalization layer (such as SegNet). I resolved the issue by using the caffe segnet version.

nurvarli commented 5 years ago

Yes, I solved this problem with caffe Segnet, too. This algorithm can find 7 landmarks default on test images. I have changed NLMK parameter with 68 but it did not work. I need 68 landmarks. Do you know, how is it possible?

vkasojhaa commented 5 years ago

I also need the pre-trained weights for 68 landmarks. I have opened an issue for the same. I guess you need to train the model on 68 landmarks but there is no code available to train the model.