Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/9144
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 55513b3 with merge base c50f694 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
| # perspective is transformed with :class:`~torchvision.transforms.RandomPerspective`. | ||
| perspective_transformer = v2.RandomPerspective(distortion_scale=0.6, p=1.0) | ||
| perspective_imgs = [perspective_transformer(orig_img, orig_box) for _ in range(4)] | ||
| plot([(orig_img, orig_box)] + perspective_imgs, bbox_width=10) |
There was a problem hiding this comment.
@scotts @AntoineSimoulin , as discussed offline with both of you separately, the results of perspective were a bit fishy, so I removed it from this tutorial. We should try to investigate what's going on and potentially provide a bug fix, if we think this is a bug.
I ralso emoved Elastic below, not because it was buggy, but because it didn't seem to transform the original image in a relevant way for the purpose of this tutorial.
| # ------- | ||
| # Rotated bounding boxes also maintain their properties when the image is padded using | ||
| # :class:`~torchvision.transforms.Pad`. | ||
| # Using # :class:`~torchvision.transforms.Pad`: |
|
Hey @NicolasHug! You merged this PR, but no labels were added. |
Reviewed By: AntoineSimoulin Differential Revision: D79175018 fbshipit-source-id: 85ff7a8bd13e847a283bd9f527280fa27fd31252
This PR adds more docs for the rotated bounding boxes, as documents them as BETA, along with the new KeyPoints.
We'll also add a note about those things being BETA in our release notes.
You can view the built docs by clicking on the link from the
pytorch-botcomment just below