Skip to content

A set of image attention layers implemented as custom keras layers that can be imported dirctly into keras

License

Notifications You must be signed in to change notification settings

NextTechLab/Visual_attention_tf

This branch is 10 commits behind vinayak19th/Visual_attention_tf:main.

Repository files navigation

Visual_attention_tf

GitHub license PyPI - Python Version PyPI PyPI - Wheel

A set of image attention layers implemented as custom keras layers that can be imported dirctly into keras

Currently Implemented layers:

Installation

You can see the projects official pypi page : https://pypi.org/project/visual-attention-tf/

pip install visual-attention-tf

Usage:

from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Conv2D
from visual_attention import PixelAttention2D , ChannelAttention2D

inp = Input(shape=(1920,1080,3))
cnn_layer = Conv2D(32,3,,activation='relu', padding='same')(inp)

# Using the .shape[-1] to simplify network modifications. Can directly input number of channels as well
Pixel_attention_cnn = PixelAttention2D(cnn_layer.shape[-1])(cnn_layer)
Channel_attention_cnn = ChannelAttention2D(cnn_layer.shape[-1])(cnn_layer)

About

A set of image attention layers implemented as custom keras layers that can be imported dirctly into keras

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%