resattention 0.0.4

A library for implementing Gated Residual Attention CNN for image classification problems.

# resattention

A python library for implementing a Residual Attention Convolutional Neural Network and training it for image classification problems. This model supports multi-class classification and is easy to use both for training and testing purposes.

The code is Python 2 and 3 compatible.

# Installation

Fast install:

pip install resattention

For a manual install get this package:

$wget https://github.com/garain/resattention/archive/master.zip
$unzip master.zip
$rm master.zip
$cd resattention-master

Install the package:

python setup.py install

# Example

from resattention import models
import tensorflow as tf

model = models.AttentionResNetCifar10(shape=(32,32,3),n_classes=2,n_channels=3)#RGB images
model = models.AttentionResNetCifar10(shape=(32,32,1),n_classes=6,n_channels=1)#GrayScale images

model.compile(tf.keras.optimizers.Adam(lr=0.0001), loss='categorical_crossentropy', metrics=['accuracy'])
model.summary()

# Models supported

• AttentionResNetCifar10

• AttentionResNet92

• AttentionResNet56

# Please cite this publication if this library comes to any use:

1. Garain, B. Ray, P. K. Singh, A. Ahmadian, N. Senu and R. Sarkar, “GRANet: A Deep Learning Model for Classification of Age and Gender from Facial Images,” in IEEE Access, doi: 10.1109/ACCESS.2021.3085971.