ham952 / segmentation-keras

Implementation of custom FCN architecture along with base FCN and UNet Architectures in Keras

Segmentation-keras

This repository includes implementation of following FCN implementations for segmentation task

1. FCN8-VGG16
2. FCN8-VGG19
3. FCN8-ResNet50

Furthermore a custom FCN architecture is also implemented refered here as FCN2 including

 1. FCN2-VGG16
 2. FCN-VGG16
 3. FCN2-ResNet50

Lastly, Unet architecture is implemnted on VGG16 as following

 1. UNet-VGG16

FCN2 architecture in comparison with FCN8 is as following

Dataset

The segmnetation data is taken from Divam Gupta available at image Segmentation git

Results

1. FCN2 Vs FCN8 Implementations

2. UNet Vs FCN Implementations

Conlcusion

• FCN2 is the natural extension of FCN architectures
• FCN2 under current experiment settings, donot provide any better performance.
• Ablation studies need to ascertain the performance gaps within the FCN variants
• Unet architecture is the most closest implmentation of proposed FCN2 architecture

Citation Details

If you find this code useful in your research, please consider citing

@misc{xbd-classification,
author = {{Rafique, Hamza}},
title = {Experimenting with xBD classification task for xview2 challenge using Keras},
year = {2019},
version = {1.0}
address = {Air University, ISB. [email protected]},
url = {https://github.com/ham952/xview2-pytorch-firstrun}
} 

ToDo

• Addition of copyrights in code files
• Improving results presentation in git