Skip to content

hamidriasat/BiSeNetV2

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 

Repository files navigation

BiSeNet V2: Bilateral Network with Guided Aggregation for Real-time Semantic Segmentation

This repository contain implementation of BiSeNet V2 in Tensorflow/Keras.

Overview

overview-of-bisenet v2 method

Detailed Implementation

overview-of-our-method
Instantiation of the Detail Branch and Semantic Branch. Each stage S contains one or more operations opr (e.g., Conv2d, Stem, GE, CE). Each operation has a kernels size k, stride s and output channels c, repeated r times. The expansion factor e is applied to expand the channel number of the operation. Here the channel ratio is ɑ = 1/4. The green colors mark fewer channels of Semantic Branch in the corresponding stage of the Detail Branch. Notation: Conv2d means the convolutional layer, followed by one batch normalization layer and relu activation function. Stem indicates the stem block. GE represents the gather-and-expansion layer. CE is the context embedding block.

Dependies:
Tensorflow 2.0 or later

Licensed under MIT License