crisie / recurrentgaze Goto Github PK

Greetings,

When I run "python OnlineTest.py" I get the following error:

Process finished with exit code 139 (interrupted by signal 11: SIGSEGV)

it occurs when

success = self.info3D_model.trackFace(image.encode('utf-8'), calib_matrix_np, pointer(face_info))

is executed.

Has anyone encountered this?
I know it is some kind of memory access problem.
I'm using Ubuntu 20.04 lts.

Thank you for your help and best regards

Hey,
I was trying to run TestOnline.py 's code and faced this issue

AttributeError                            Traceback (most recent call last)
<ipython-input-8-b74b6227f978> in <module>
      6 
      7 # Load models - change to device = 'gpu' to run with cuda support
----> 8 ga = gaze_estimation.GazeEstimation(device='gpu', info3D_lib_path=openfacewrapper_path)
      9 
     10 dir = os.path.dirname(__file__)

~/brainalive/RecurrentGaze/gaze_estimation.py in __init__(self, model_type, device, info3D_lib_path)
     48 
     49         self.batch_size = 1
---> 50         self.landmarks_model = LandmarksEstimation(device=device).get_model()
     51         self.gaze_model = ExperimentHelper.get_experiment(model_type.value)
     52         self.info3D_model = Info3DEstimation(lib_path=info3D_lib_path)

~/brainalive/RecurrentGaze/landmarks_estimation.py in get_model(cls)
     18     @classmethod
     19     def get_model(cls):
---> 20         return LandmarksEstimation.landmarks_model

AttributeError: type object 'LandmarksEstimation' has no attribute 'landmarks_model'

I have followed the steps in the README.md and provided the path to libOpenFaceWrapper.so file in openfacewrapper_path as "../OpenFaceWrapper/build"
Any help is appreciated
Edit :
When trying to run
face_alignment.FaceAlignment(model_type, flip_input=False, device='cuda')
separately It runs with no issues, and I have copies the AU_predictors and model folder to the script's location
Edit 2:
I using Ubuntu 16.04 and Test.py is working fine

if I can test the model to estimate the gaze only input one frame?

Hello,

Nice work! I just have a few questions regarding the generate_training_X file.

It seems like it expects a "frames" folder that contains a few files that we need to generate by ourselves for the EYEDIAP dataset:

1. extracted bmp images from the original videos (the rgb_vga.mov one? )
2. 3D landmarks using Bulat et al. 2017 code, that should be formatted as 'result3D_' + session_str + '.txt', correct?

Thanks!

Hello! I'd like to try RecurrentGaze out for my project. Do you have pre-trained weights saved? If you have - I'd be grateful if you share them too.

Hi Sir,
Could you please share a pretrained model? The EYEDIAP dataset is only shared to those guys in school. I can not train a model from scratch.

Thanks very much.

I followed this step and it run correctly.
https://github.com/crisie/RecurrentGaze#test

But it doesn't work with new images.
It seems like I should have other files.
How can I make these files? ("CCS_3D_info.txt", "landmarks.txt", "calibration.txt")

It looks like there are no code to make the files for inference.
In README.md, there are only descriptions that you used Openface.
Please tell me about the format of 3 files or how to make that.

Hi,

I was trying to do some gaze estimation test on a simple image. I have openface compiled.

When I run it, the error says:
/OpenFace-master/OpenFaceDLL.so: cannot open shared object file: No such file or directory

What exactly is this OpenFaceDLL.so file?