Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Augmentation pipeline block #2

Closed
11 tasks done
talmo opened this issue Jul 5, 2023 · 0 comments · Fixed by #12
Closed
11 tasks done

Augmentation pipeline block #2

talmo opened this issue Jul 5, 2023 · 0 comments · Fixed by #12

Comments

@talmo
Copy link
Contributor

talmo commented Jul 5, 2023

Block/module for flexible augmentation. This should be performant and flexible with respect to the specific augmentations so we can add more in the future.

  • Implement backend(s)
    • Explore using Kornia if possible to accelerate operations on GPU
    • Also support Albumentations (CPU-only?)
  • Make sure (and add tests) for ensuring that augmentations are applied to both the image and the keypoints!
  • Implement core augmentations
    • Rotations
    • Scaling
  • Implement additional augmentations
    • Contrast/brightness
    • Cropping
    • Gaussian blur
    • Dropout patches
    • MixUp

Notes:

  • Topdown pipelines that operate on instance-level examples may benefit from specialized augmentation (see Top-down centered-instance pipeline #3 notes)
  • Implement frame-level augmentation to start for generality
  • If possible, try to keep it general enough to handle both single frame and multi-frame by operating on batches (where single-frame is special case of batch size = 1). This will make it easier to generalize to multi-frame pipelines later.
This was referenced Jul 5, 2023
@alckasoc alckasoc linked a pull request Jul 27, 2023 that will close this issue
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging a pull request may close this issue.

1 participant