Skip to content

Extend the supported types of decodePNG#2984

Merged
fmassa merged 4 commits intopytorch:masterfrom
datumbox:feature/read_image_pngtypes
Nov 10, 2020
Merged

Extend the supported types of decodePNG#2984
fmassa merged 4 commits intopytorch:masterfrom
datumbox:feature/read_image_pngtypes

Conversation

@datumbox
Copy link
Copy Markdown
Contributor

@datumbox datumbox commented Nov 10, 2020 •
edited
Loading

Fixes #2823

This PR adds support of RGB with transparency, grayscale, grayscale with transparency and palette to the decodePNG() method. To test the changes we reuse the test_decode_png() unit-test and we add additional data. The test compares for differences between reading data from PIL vs torchvision.io.image.

The following script was used to produce the additional images required for the test:

from PIL import Image

# manually downloaded from https://pytorch.org/assets/images/pytorch-logo.png
original = 'pytorch-logo.png'
path = 'rgbalpha_pytorch.png'
with Image.open(original) as img:
    img = img.convert("RGBA")
    img = img.resize((100, 100)) 
    img.save(path)

with Image.open(path) as img:
    img = img.convert("P", palette=Image.ADAPTIVE, colors=32) 
    img.save("pallete_pytorch.png")

with Image.open(path) as img:
    img = img.convert("L") 
    img.save("gray_pytorch.png")

with Image.open(path) as img:
    img = img.convert("LA") 
    img.save("grayalpha_pytorch.png")

with Image.open(path) as img:
    img = img.convert("RGB") 
    img.save("rgb_pytorch.png")

Notes:

  • To avoid any licensing issues, we use the pytorch logo.
  • The images are shrunk to 100x100 pixels to keep their total size less than 4kb.
  • The use of real images allows us to reuse existing tests. It also allows for easier manual inspection/visualization.

fmassa
fmassa approved these changes Nov 10, 2020
View reviewed changes
Copy link
Copy Markdown
Member

@fmassa fmassa left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks great, thanks a lot!

@fmassa fmassa merged commit 78159d6 into pytorch:master Nov 10, 2020
@datumbox datumbox deleted the feature/read_image_pngtypes branch November 10, 2020 15:12
@datumbox datumbox mentioned this pull request Nov 10, 2020
Merged
bryant1410 pushed a commit to bryant1410/vision-1 that referenced this pull request Nov 22, 2020
@datumbox @bryant1410
Extend the supported types of decodePNG (pytorch#2984)
6026fea 
* Add support of different color types in readpng.

* Adding test images and unit-tests.

* Use closest possible type.

* Fix formatting.
@datumbox datumbox mentioned this pull request Dec 1, 2020
Merged
vfdev-5 pushed a commit to Quansight/vision that referenced this pull request Dec 4, 2020
@datumbox @vfdev-5
Extend the supported types of decodePNG (pytorch#2984)
e9190b2 
* Add support of different color types in readpng.

* Adding test images and unit-tests.

* Use closest possible type.

* Fix formatting.
@datumbox datumbox mentioned this pull request Jan 5, 2021
Closed
13 tasks
@datumbox datumbox mentioned this pull request Feb 1, 2021
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@fmassa fmassa fmassa approved these changes

Assignees

No one assigned

Labels

cla signed

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Support paletted images for io.image.read_image

3 participants

@datumbox @fmassa @facebook-github-bot