pytorch / pytorch/vision

Return only images from `torchvision.dataset` datasets

Open
#5,601 3 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

🚀 The feature

torchvision.dataset datasets should provide an option (e.g., a constructor argument) to return only the image (i.e., rather than an image and target pair).

Motivation, pitch

I work on compression research and would like to use torchvision.dataset datasets but, as of now, they are too cumbersome to use because they provide more input than what is needed by my models (i.e., images and targets rather than just images).

Alternatives

It is entirely possible a straightforward alternative exists! If so, I’d love to read about it!

Additional context

I suspect this would also be welcome by any other community working on image to image methods.

cc @pmeier

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

Start from the torchvision.dataset dataset constructors and inspect how they currently return image-target pairs. Determine the scope of the image-only option across datasets, then verify that image-only iteration works while existing pair-returning behavior remains unchanged.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
computer-vision, machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.