Skip to content

A Tensorflow 2+ implementation of the DeepCrack segmentation model

License

Notifications You must be signed in to change notification settings

DavidHidde/DeepCrack-Tensorflow

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DeepCrack-Tensorflow

This is a simple Tensorflow 2+ implementation of the DeepCrack segmentation model.
This is a Tensorflow translation of the original PyTorch implementation using the TensorFlow functional API. The implementation has been verified using TensorFlow 2.16.

There are some minor differences between the PyTorch code, the paper and this implementation. Compared to the PyTorch implementation, batch normalization layers have been added after each convolution since this was indicated in the paper and SegNet. Compared to the paper, this implementation and the original PyTorch implementation implement skip layers using a 64-filter convolution, 2D upsampling and then a 1-filter convolution. From empirical testing, this has shown to improve the loss minimization of the model.

Disclaimer

The code in this repository is based on:

For older Tensorflow versions, hanshenChen/crack-detection is recommended.

About

A Tensorflow 2+ implementation of the DeepCrack segmentation model

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages