Converting .t7 labels into other formats about 2d-and-3d-face-alignment HOT 3 CLOSED

Hi @azmathmoosa ,

Are you using Windows? If yes, then there seems to be an issue with the library you are trying to use, see bshillingford/python-torchfile#9

Regarding an utility to do this, you can use pytorch (there is a windows build for it using conda). If this still doesn't work, please let me know and I will have a look.

from 2d-and-3d-face-alignment.

Thanks for the quick response @1adrianb . I tried pytorch for windows and it too failed to deserialize.

>>> import torch
>>> from torch.utils.serialization import load_lua
>>>
>>> a = load_lua('aflw__face_40662.t7')
>>> a
[torch.DoubleTensor with no dimension]

Going through the issue that you referenced they say that binary t7 files are not portable due to different sizes of long ints. So I must use a linux system to deserialize them first and then switch to windows. I will do that then.

Thanks for the assistance, and great work!

from 2d-and-3d-face-alignment.

I have finished the conversion, Win10, python3.7, pytorch 0.4.1(cpu)
use code

import torch
from torch.utils.serialization import load_lua
# code...
t7_data = load_lua("the path", long_size=8)
# code...

from 2d-and-3d-face-alignment.