Adrian Bulat

The weights of each group are identical and they simply reflect the internal changes made inside pytorch. Beyond this 2DFAN will produce 2D landmarks, 3DFAN - 2.5D landmarks and 3DFAN...

Sorry for delay, yes, but I don't have an exact timeline unfortunately.

@LinusLGL It looks like something was not installed correctly or the package is at on older version. The first error implies that the package installation failed. Install it by running...

To followup on this, the class does contain that argument (see https://github.com/1adrianb/face-alignment/blob/c49ca6fef8ffa95a0ac7ce698e0b752ac91f6d42/face_alignment/api.py#L53). Please make sure you are using the latest version (1.3.4). You can check this using ```face_alignment.__version__```

@LinusLGL ```pip install face-alignment -U```

The resolutions is not fixed. The code will automatically detect the faces and pre-process the images as needed. The larger the images - the more time it will take thought....

@miracleyoo if you have time, feel free to make a pull request :)

I just run the .py file from the examples folder and I was unable to reproduce your error, neither on a windows nor linux machine. Please make sure that you...

Hi @Neltherion, This should definitely alleviate the issue to some extend. The only reason I didn't do this is because the face detector used (SFD) can be quite memory intensive...

Sure, we can add as an optional paramter the batch size, with the default set to 1