https://github.com/titu1994/Keras-ResNeXt

Implementation of ResNeXt models from the paper Aggregated Residual Transformations for Deep Neural Networks in Keras 2.0+.
https://github.com/titu1994/Keras-ResNeXt

Last synced: 3 days ago
JSON representation

Implementation of ResNeXt models from the paper Aggregated Residual Transformations for Deep Neural Networks in Keras 2.0+.

• Host: GitHub
• URL: https://github.com/titu1994/Keras-ResNeXt
• Owner: titu1994
• License: mit
• Created: 2017-06-09T16:25:11.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2021-04-17T10:05:32.000Z (over 3 years ago)
• Last Synced: 2024-10-26T21:34:10.946Z (8 days ago)
• Language: Python
• Homepage:
• Size: 630 KB
• Stars: 224
• Watchers: 9
• Forks: 80
• Open Issues: 8
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-image-classification - unofficial-keras : https://github.com/titu1994/Keras-ResNeXt
• awesome-image-classification - unofficial-keras : https://github.com/titu1994/Keras-ResNeXt

README

        
# Keras ResNeXt

Implementation of ResNeXt models from the paper [Aggregated Residual Transformations for Deep Neural Networks](https://arxiv.org/pdf/1611.05431.pdf) in Keras 2.0+.

Contains code for building the general ResNeXt model (optimized for datasets similar to CIFAR) and ResNeXtImageNet (optimized for the ImageNet dataset).

# Salient Features

ResNeXt updates the ResNet block with a new expanded block architecture, which depends on the `cardinality` parameter. It can be further visualised in the below diagram from the paper.

![cardinality](https://github.com/titu1994/Keras-ResNeXt/blob/master/images/Cardinality.PNG?raw=true)

---

However, since grouped convolutions are not directly available in Keras, an equivalent variant is used in this repository (see block 2)

![equivalent blocks](https://github.com/titu1994/Keras-ResNeXt/blob/master/images/equivalent_blocks.PNG?raw=true)

# Usage

For the general ResNeXt model (for all datasets other than ImageNet),

```

from resnext import ResNext

model = ResNext(image_shape, depth, cardinality, width, weight_decay)

```

For the ResNeXt model which has been optimized for ImageNet,

```

from resnext import ResNextImageNet

image_shape = (112, 112, 3) if K.image_data_format() == 'channels_last' else (3, 112, 112)

model = ResNextImageNet(image_shape)

```

Note, there are other parameters such as depth, cardinality, width and weight_decay just as in the general model, however the defaults are set according to the paper.