targetran - Targetran 0.13.2

• Fixed some NumPy 2.0 compatibility and mypy related issues. No changes to logic.

- Python
Published by bhky over 1 year ago

targetran - Targetran 0.13.1

• Default empty tuple for the bboxes_seq and labels_seq args of to_tf and to_tf_dataset functions, so that it will be a bit less cumbersome dealing with image-only cases.

- Python
Published by bhky over 1 year ago

targetran - Targetran 0.13.0

• Drop support for Python 3.8; added support for 3.11.
• Rename seqs_to_tf_dataset to to_tf_dataset.
• Add keras-cv into requirements.txt.
• Add new functions to_keras_cv_dict and to_keras_cv_model_input.
• Update docs.

- Python
Published by bhky almost 2 years ago

targetran - Targetran 0.12.1

• Fix logic in targetran.utils.Compose.

- Python
Published by bhky over 2 years ago

targetran - Targetran 0.12.0

• Add fill_value option for assigning value to be used to fill the out-of-boundaries regions, in the past it must be 0
• Add keep_order option in TFCombineAffine and CombineAffine for ensuring the (randomly) chosen transformation steps will be executed in the provided order
• More interpolation options for image resizing
• Fix transpose bug

- Python
Published by bhky almost 4 years ago

targetran - Targetran 0.11.8

• Fixed bug that created one-pixel shift for some transformation cases
• Simplified logic in base affine transformation function

- Python
Published by bhky almost 4 years ago

targetran - Targetran 0.11.7

• Fix range check: lower limit should be inclusive, and upper limit is exclusive if its value differs from the lower one

- Python
Published by bhky almost 4 years ago

targetran - Targetran 0.10.4

• Python 3.7 only
• Fix range check: lower limit should be inclusive, and upper limit is exclusive if its value differs from the lower one

- Python
Published by bhky almost 4 years ago

targetran - Targetran 0.11.6

• Fix type annotations

- Python
Published by bhky almost 4 years ago

targetran - Targetran 0.10.3

• Python 3.7 only
• Fix type annotations

- Python
Published by bhky almost 4 years ago

targetran - Targetran 0.11.5

Fix instantiating TF random number generator: do not use from non deterministic state

- Python
Published by bhky about 4 years ago

targetran - Targetran 0.10.2

• Python 3.7 only
• Fix instantiating TF random number generator: do not use from non deterministic state

- Python
Published by bhky about 4 years ago

targetran - Targetran 0.11.4

Fix typing issues

- Python
Published by bhky about 4 years ago

targetran - Targetran 0.11.3

Relax TF version requirement.

- Python
Published by bhky about 4 years ago

targetran - Targetran 0.11.1

Minor fix of internal methods

- Python
Published by bhky about 4 years ago

targetran - Targetran 0.11.0

• Fix typing issues
• Drop support for Python 3.7, as NumPy has stopped supporting it since 1.22.0
• Set minimum required NumPy version as 1.22.0

- Python
Published by bhky about 4 years ago

targetran - Targetran 0.10.0

Breaking changes: - Use newest numpy.typing feature to strengthen the type annotations of the NumPy related code, see here for the notations.

- Python
Published by bhky over 4 years ago

targetran - Targetran 0.9.3

Update information dunders.

- Python
Published by bhky over 4 years ago

targetran - Targetran 0.9.2

Python requirement updated. It now requires python>=3.7.

- Python
Published by bhky over 4 years ago

targetran - Targetran 0.9.1

Improve code speed for BILINEAR interpolation.

- Python
Published by bhky over 4 years ago

targetran - Targetran 0.9.0

Add interpolation mode option to: - CombineAffine/TFCombineAffine - RandomRotate/TFRandomRotate/rotate/tf_rotate - RandomShear/TFRandomShear/shear/tf_shear - RandomTranslate/TFRandomTranslate/translate/tf_translate

Currently, BILINEAR (default) and NEAREST are supported.

- Python
Published by bhky over 4 years ago

targetran - Targetran 0.8.2

One can now do ds = seqs_to_tf_dataset(images, [], []) to get a Dataset which returns (image, [], []) on each iteration. This is more convenient than creating dummy bboxes and labels when working with image data only.

- Python
Published by bhky over 4 years ago

targetran - Targetran 0.8.1

Fix getting random transform steps in TFCombineAffine.

- Python
Published by bhky over 4 years ago

targetran - Targetran 0.8.0

Add support for randomly selecting only a subset of transform steps to use, with new parameters num_selected_transforms and selected_probabilites introduced to CombineAffine and TFCombineAffine.

- Python
Published by bhky over 4 years ago

targetran - Targetran 0.7.7

Change the default probability of the random flip to 0.5

- Python
Published by bhky almost 5 years ago

targetran - Version 0.7.6

• Default probability of CombineAffine and TFCombineAffine is changed to 1.0
• Default probability of other random transformation classes is changed to 0.9

- Python
Published by bhky almost 5 years ago

targetran - Targetran 0.7.5

Change default values of some parameters - random rotate angle range: (-15.0, 15.0) - random shear angle range: (-10.0, 10.0) - random translate height/width range: (-0.1, 0.1) - random crop height/width range: (0.8, 0.9)

- Python
Published by bhky almost 5 years ago

targetran - Targetran 0.7.4

Breaking changes: shorten some input parameter names regarding cropping, e.g., cropped_image_height -> crop_height.

- Python
Published by bhky almost 5 years ago

targetran - Targetran 0.7.3

Initial public release of Targetran, a data augmentation library to assist object detection or image classification model training. It can be used with TensorFlow or PyTorch.

- Python
Published by bhky almost 5 years ago