KeyError: 'content-type' when extracting face with dlib

[hungpham3112]

Today is the first time I use this package, I want to extract the face from the image.

Here is my code:
detected_face = DeepFace.extract_faces("./obama.jpg", detector_backend = 'dlib')

But then the issue appeared:

It seems that deepface was trying to download the model but it can't

OS: Pop-OS 22.04
Env: conda
Editor: jupyter notebook

[serengil]

This is covered here: https://github.com/serengil/deepface/blob/master/deepface/basemodels/DlibResNet.py#LL30C21-L30C43

And the target url is http://dlib.net/files/dlib_face_recognition_resnet_model_v1.dat.bz2

It is still alive. It seems you cannot access that url. Check your firewall or proxy.

[hungpham3112]

This is covered here: https://github.com/serengil/deepface/blob/master/deepface/basemodels/DlibResNet.py#LL30C21-L30C43

And the target url is http://dlib.net/files/dlib_face_recognition_resnet_model_v1.dat.bz2

It is still alive. It seems you cannot access that url. Check your firewall or proxy.

The problem is http, I downloaded with firefox and it blocked. Is there any way to handle that with Python?

[serengil]

This is not happening in my env. Sorry.

[hungpham3112]

This is not happening in my env. Sorry.

Can you provide more information about your environment? Are you using Linux or Windows?

[serengil]

Tested it on mac & debian.

[hungpham3112]

I found the problem and can fix this with PR

[serengil]

PRs are of course welcome but this is because of your env, not a bug.

[kennedyjosh]

I am also experiencing this issue on MacOS. I am able to download from the link in my browser, but not through the code. The PR from @hungpham3112 fixed the issue – what is the harm in accepting his approach, as this 'Content-Type' field does not seem to be guaranteed?

[hungpham3112]

@kennedyjosh Can you look at the PR wkentaro/gdown#302 if it solve downloading problem?