Thank you so much for this wonderful work! In order to re-run your experiments, we need to extract the keypoints and the segmentation; however, it is not clear how to do it.
In the infer.py file, there is a sketch on how to generally extract the keypoints and the segmentation using the preprocess_data function, but some of the operations on the keypoints and segmentations in infer.py have overlap with the operations in voxceleb2.py file which is called during the training, thus the preprocess before training to save the keypoints and segmentations is different from infer.py. Furthermore, it's not clear if you have used the crop_data in the training process.
I will be most grateful if you provide us with the necessary code to generate the keypoints and segmentations from videos faithful to your design.
Dear author,
Thank you so much for this wonderful work! In order to re-run your experiments, we need to extract the keypoints and the segmentation; however, it is not clear how to do it.
In the
infer.pyfile, there is a sketch on how to generally extract the keypoints and the segmentation using thepreprocess_datafunction, but some of the operations on the keypoints and segmentations ininfer.pyhave overlap with the operations invoxceleb2.pyfile which is called during the training, thus the preprocess before training to save the keypoints and segmentations is different frominfer.py. Furthermore, it's not clear if you have used thecrop_datain the training process.I will be most grateful if you provide us with the necessary code to generate the keypoints and segmentations from videos faithful to your design.
Thanks