yinguobing / face-mesh-generator

Generate face mesh dataset using Google's FaceMesh model.
GNU General Public License v3.0
118 stars 22 forks source link

augmented dataset #7

Open soans1994 opened 3 years ago

soans1994 commented 3 years ago

hello author,

i have some queries regarding your implementation.

  1. i see that it adds extra keypoints to the existing dataset. how can i use the original landmarks.
  2. the tf record genearted does augmentation? how many augment samples?
  3. how to generate train and test split tfrecords? thank you
yinguobing commented 3 years ago

Hi @soans1994

  1. I think you can use branch features/tfrecord-marks-68 which use the original marks.
  2. The only augmentation in this branch is random minor scaling of face areas. However this won't generate new samples. You will have the same number of samples as the original dataset.
  3. You can split the samples however you want to. The whole process is wrapped in a for loop. It shouldn't be hard.
soans1994 commented 3 years ago

@yinguobing thank you for the original 68 keypoints code. i want to generate augmented images and landmarks. can you please help

yinguobing commented 3 years ago

Usually the augmentation is done online (during training). Here is an example: https://github.com/yinguobing/facial-landmark-detection-hrnet/blob/18fc4e504b41cc322a9ff7dc483dc6dfbb97aefc/dataset.py#L39

soans1994 commented 3 years ago

thank you so much, i will check it out