https://github.com/titu1994/keras-attention-augmented-convs

Keras implementation of Attention Augmented Convolutional Neural Networks
https://github.com/titu1994/keras-attention-augmented-convs

attention-augmented-conv keras-tensorflow tensorflow

Last synced: 18 days ago
JSON representation

Keras implementation of Attention Augmented Convolutional Neural Networks

• Host: GitHub
• URL: https://github.com/titu1994/keras-attention-augmented-convs
• Owner: titu1994
• License: mit
• Created: 2019-04-28T16:37:35.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2020-03-06T16:03:01.000Z (over 4 years ago)
• Last Synced: 2024-10-13T14:17:04.277Z (about 1 month ago)
• Topics: attention-augmented-conv, keras-tensorflow, tensorflow
• Language: Python
• Size: 250 KB
• Stars: 120
• Watchers: 6
• Forks: 38
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Keras Attention Augmented Convolutions

A Keras (Tensorflow only) wrapper over the Attention Augmentation module from the paper [Attention Augmented Convolutional Networks](https://arxiv.org/abs/1904.09925).

Provides a Layer for Attention Augmentation as well as a callable function to build a augmented convolution block.



# Usage

It is advisable to use the `augmented_conv2d(...)` function directly to build an attention augmented convolution block.

```python

from attn_augconv import augmented_conv2d

ip = Input(...)

x = augmented_conv2d(ip, ...)

...

```

If you wish to add the attention module seperately, you can do so using the `AttentionAugmentation1D` layer as well.

```python

from attn_augconv import AttentionAugmentation1D

ip = Input(...)

# make sure that input to the AttentionAugmentation1D layer has (2 * depth_k + depth_v) filters.

x = Conv2D(2 * depth_k + depth_v, ...)(ip)

x = AttentionAugmentation1D(depth_k, depth_v, num_heads)(x)

...

```

# Requirements

  - Tensorflow 2.0+