pytorch / pytorch/vision

Supporting "seasonal" transforms for image augmentation in torchvision

Open
#7,633 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
17.9k
Forks
7.3k
Avg merge
1d 15h
Merged PRs (30d)
13

Description

🚀 The feature

I was wondering if there's a plan to add seasonal transforms for real-world iaug to TV transforms. For example: https://github.com/UjjwalSaxena/Automold--Road-Augmentation-Library includes:

  1. Fog
  2. Sun Flare
  3. Snow
  4. Shadows (typically from buildings nearby)

Again motivated by the fact that these are things one expects to see in the real-world, and Albumentations includes those.

High level Q. Is there a plan to be somewhat transform compatible with Albumentations?

Follow-up from https://github.com/pytorch/vision/issues/7620#issuecomment-1562413892

Motivation, pitch

"Seasonal" transforms are useful for tasks that potentially involve real-world data outdoors.

Alternatives

No response

Additional context

Following up from https://github.com/pytorch/vision/issues/7620#issuecomment-1562413892

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Review the torchvision transforms area, the linked follow-up discussion in issue #7620, and the referenced Automold and Albumentations implementations. Done would require a decided scope and acceptance criteria for seasonal transforms, but this issue names no files, tests, or bounded first change.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-vision, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.