Alexandre Tuleu

Some projects requires a specific clang version to compiles. This dirty patch allows to use cldoc with them.

such option should : - [ ] imply the `-pedantic` flag, see #63 - [ ] instead of reporting error to stderr, it should prompt user if he want to...

Such option : - [ ] be disabled by default - [ ] would turn any warning into error - [ ] would turn any undocumented entity found into error...

A common mistake when cldoc-ing a document is to leave the first comment line empty. ``` c++ /* * A does */ class A{ public: void Thing(); }; ``` This...

**System information** - Pop_OS 22.04 - Not installed - Not installed (source build) - Python version: 3.8.12 - Is GPU used? (yes/no): yes **Describe the bug** While trying a source...

Towards keypoint support: Adds no-op operations some on existing augmentation

9

# What does this PR do? Adds keypoints bounding_boxes no-op operations. It does so by adding `with_keypoints.py` and `with_bounding_boxes.py` tests that ensure all augmentation supports these augmentations Related to #518...

Fixes `label_smoothing` in `FocalLoss`

1

# What does this PR do? Fixes label smoothing in `FocalLoss`. Also as a code style, uses the same formulation for `alpha` and `pt` and computation Fixes #672 ## Before...

Adds different class reduction scheme in `FocalLoss`

3

**Short Description** Currently `FocalLoss` reduces multi - label problems by a `mean` (see below): https://github.com/keras-team/keras-cv/blob/ff20079aebe8e80bc355c279c9c5541a7c823535/keras_cv/losses/focal.py#L90 However a strictly correct formulation would be a `sum`. Original implementaion ( Facebook Research, pytorch...

Possibly wrong implementation of focal loss in regard of label smoothing

10

I think that there is a wrong implementation of focal loss in regard of label smoothing. Here is the current implementation https://github.com/keras-team/keras-cv/blob/35b9f9534fc193e00b4fa850963a05b331467065/keras_cv/losses/focal.py#L74-L90 as seen above, label smoothing occurs *before* the...

Toward keypoint supports: Adds a `bounding_box.transform_from_corners_fn` function

1

# What does this PR do? Adds a function that will be used as a generic way of approximating bounding boxes augmentations by applying keypoints augmentation on each 4 corners...