danduncan / happynet Goto Github PK

I follow the steps you provide strictly, but the loss is -nan from beginning. The training data is properly converted, what is the reason?

Hi, I am new to caffe. When I searched for the caffe model on the Github, they have several version of caffe model. Such as EmotiW with RGB, EmotiW with LBP etc. Which version of Caffe model I should choose?
Thanks!

Hi ,i training my own datasets but the accuracy is only 25%,what should i do to improve the accuracy

When I convert ck dataset to jaffe format, i have this issue:

Detected 1 faces.
Traceback (most recent call last):
File "D:\software\work\Caffe\models\HappyNet-master\HappyNet-master\CKPlus_convert_to_Jaffe_format.py", line 64, in
input_list = faceCrop(dirCrop, input_list, color, single_face)
File "D:\software\work\Caffe\models\HappyNet-master\HappyNet-master\opencv_functions.py", line 212, in faceCrop
cropped_pil_img.save(save_name)
File "D:\software\work\Anaconda2\lib\site-packages\PIL\Image.py", line 1725, in save
fp = builtins.open(filename, "w+b")
IOError: [Errno 2] No such file or directory: 'datasets/CK_Plus/cropped/CKPlus_Images\S005\001\S005_001_00000011_crop1.png'
[Finished in 2.8s]

i can not understand why this happen, thank you very mucn for help!

Hi there!
Seems like you missed some files - would be nice if you could push them too:

    video_generate.py  - Run HappyNet in real-time and save output to video
    video_test.py      - Run HappyNet in real-time; does not save to video

Thanks in advance!