pytorch / pytorch/vision

[feature proposal] TwentyCrop 💥

Open
#877 0 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

Hi,

I decided to implement TwentyCrop for one of my project. Compared to TenCrop, it flips image BOTH vertically AND horizontally. 💥
The implementation is very easy starting from TenCrop's code! You can see it here (Google Colab) or here (GitHub) (search for TwentyCrop).

For this project, I have trained multiple models, and have tested them with 1, 5, 10 and 20-crop.
The best scores can be found here (Google Colab).

Theses "statistics" show that (at least on my particular project), almost half the time 20-crop does better than 10-crop, so (again at least on my particular project) TwentyCrop is worthwhile.

Do you think it's worthwile to implement TwentyCrop in PyTorch? If yes, I can fork it and make a pull request.

cc @vfdev-5

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 by locating PyTorch Vision's existing TenCrop implementation and compare it with the linked Google Colab or GitHub notebook's TwentyCrop example. Done means adding the proposed TwentyCrop transform with both vertical and horizontal flipping behavior; the issue does not name a test file.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
computer-vision
Issue type
Feature
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.