https://github.com/titu1994/keras-global-context-networks

Keras implementation of Global Context Attention blocks
https://github.com/titu1994/keras-global-context-networks

global-context keras tensorflow

Last synced: 9 days ago
JSON representation

Keras implementation of Global Context Attention blocks

• Host: GitHub
• URL: https://github.com/titu1994/keras-global-context-networks
• Owner: titu1994
• License: mit
• Created: 2019-04-28T18:52:42.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2019-04-29T00:47:03.000Z (over 5 years ago)
• Last Synced: 2024-10-13T14:17:11.722Z (24 days ago)
• Topics: global-context, keras, tensorflow
• Language: Python
• Size: 351 KB
• Stars: 46
• Watchers: 7
• Forks: 12
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Keras Global Context Attention Blocks

Keras implementation of the Global Context block from the paper [GCNet: Non-local Networks Meet Squeeze-Excitation Networks and Beyond](https://arxiv.org/abs/1904.11492).

Supports Conv1D, Conv2D and Conv3D directly with no modifications.



# Usage

Import `global_context_block` from `gc.py` and provide it a tensor as input.

```python

from gc import global_context_block

ip = Input(...)

x = ConvND(...)(ip)

# apply Global Context

x = global_context_block(x, reduction_ratio=16, transform_activation='linear')

...

```

# Parameters

There are just two parameters to manage : 

```

 - reduction_ratio: The ratio to scale the transform block.

 - transform_activation: The activation function prior to addition of the input with the context.

                         The paper uses no activation, but `sigmoid` may do better.

```

# Requirements

  - Keras 2.2.4+

  - Tensorflow (1.13+) or CNTK