https://github.com/moskomule/dda

Differentiable Data Augmentation Library
https://github.com/moskomule/dda

data-augmentation differentiable-data-augmentation faster-auto-augment pytorch

Last synced: about 2 months ago
JSON representation

Differentiable Data Augmentation Library

• Host: GitHub
• URL: https://github.com/moskomule/dda
• Owner: moskomule
• License: mit
• Created: 2020-02-29T11:25:44.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2022-08-07T06:28:56.000Z (over 2 years ago)
• Last Synced: 2024-11-15T05:32:09.858Z (about 2 months ago)
• Topics: data-augmentation, differentiable-data-augmentation, faster-auto-augment, pytorch
• Language: Jupyter Notebook
• Homepage:
• Size: 2.67 MB
• Stars: 120
• Watchers: 7
• Forks: 12
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Differentiable Data Augmentation Library

![](https://github.com/moskomule/dda/workflows/pytest/badge.svg)

This library is a core of Faster AutoAugment and its descendants. This library is research oriented, and its AIP may change in the near future.

## Requirements and Installation

### Requirements

```

Python>=3.8

PyTorch>=1.5.0

torchvision>=0.6

kornia>=0.2

```

### Installation

```

pip install -U git+https://github.com/moskomule/dda

```

## APIs

### `dda.functional`

Basic operations that can be differentiable w.r.t. the magnitude parameter `mag`. When `mag=0`, no augmentation is applied, and when `mag=1` (and `mag=-1` if it exists), the severest augmentation is applied. As introduced in Faster AutoAugment, some operations use straight-through estimator to be differentiated w.r.t. their magnitude parameters.

```python

def operation(img: torch.Tensor,

              mag: Optional[torch.Tensor]) -> torch.Tensor:

    ...

```

`dda.pil` contains the similar APIs using PIL (not differentiable).

### `dda.operations`

```python

class Operation(nn.Module):

   

    def __init__(self,

                 initial_magnitude: Optional[float] = None,

                 initial_probability: float = 0.5,

                 magnitude_range: Optional[Tuple[float, float]] = None,

                 probability_range: Optional[Tuple[float, float]] = None,

                 temperature: float = 0.1,

                 flip_magnitude: bool = False,

                 magnitude_scale: float = 1,

                 debug: bool = False):

        ...

```

If `magnitude_range=None`, `probability_range=None`, then `magnitude`, `probability` is not Parameter but Buffer, respectively.

`magnitude` moves in `magnitude_scale * magnitude_range`. 

For example, `dda.operations.Rotation` has `magnitude_range=[0, 1]` and `magnitude_scale=30` so that magnitude is between 0 to 30 degrees. 

To differentiate w.r.t. the probability parameter, `RelaxedBernoulli` is used.

## Examples

* [Faster AutoAugment](./faster_autoaugment)

* [RandAugment](./examples)

## Citation

`dda` (except RandAugment) is developed as a core library of the following research projects. 

If you use `dda` in your academic research, please cite `hataya2020a`.

```bibtex

@inproceesings{hataya2020a,

    title={{Faster AutoAugment: Learning Augmentation Strategies using Backpropagation}},

    author={Ryuichiro Hataya and Jan Zdenek and Kazuki Yoshizoe and Hideki Nakayama},

    year={2020},

    booktitle={ECCV}

}

...

```