Fail to load provided Meso pretrained model. about faceforensics HOT 2 OPEN

When I loaded provided Meso pretrained model, I got this:

Traceback (most recent call last):
  File "detect_from_image.py", line 182, in <module>
    prob = test_full_image_network(image_path, args.model_path, True)
  File "detect_from_image.py", line 129, in test_full_image_network
    model = torch.load(model_path)
  File "/data/***/miniconda3/lib/python3.6/site-packages/torch/serialization.py", line 386, in load
    return _load(f, map_location, pickle_module, **pickle_load_args)
  File "/data/***/miniconda3/lib/python3.6/site-packages/torch/serialization.py", line 563, in _load
    magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, 'H'.

The Meso pretrained model is downloaded from provided link in this repo and the detect_from_image.py is obtained by modifying original detect_from_video.py.

I think this is because the Pytorch only supports weight file such as .pth, the .h5 weight file is used in Caffe and Tensorflow. So what should I do to solve this problem? Have some other test shell to use?

Thanks~:)

Hello, Daisy. I think you can load the pre-trained model here provided by Andreas and go back to the author(Afchar) of Mesonet to use the original evaluation pipeline there.

from faceforensics.

Hello, Daisy.I wonder if you solved this question?I want to use the pre-trained model of face detector but have not found one, can u give some solutions?

from faceforensics.