Morphological Layers for Keras/Tensorflow2 The goal of morpholayers is to make the interactions between mathematical morphology and deep learning accessible for everyone.
If you find this code useful in your research, please consider citing:
@inproceedings{VelascoBMVC2022,
Author = {Velasco-Forero, S. and Rhim, A. and Angulo, J.},
Title = {Fixed Point Layers for Geodesic Morphological Operations},
Booktitle = {British Machine Vision Conference (BMVC)},
Year = {2022}
}
@article{VelascoSIAM2022,
author = {Velasco-Forero, Santiago and Pag\`{e}s, R. and Angulo, Jesus},
title = {Learnable Empirical Mode Decomposition based on Mathematical Morphology},
journal = {SIAM Journal on Imaging Sciences},
volume = {15},
number = {1},
pages = {23-44},
year = {2022},
}
Several examples of this library are available at: Examples
ECSIA mini-cours (Mathematical morphology meets Deep Learning)
Santiago VELASCO-FORERO, Samy BLUSSEAU, Mateus SANGALLI
Centre de Morphologie Mathématique
MINES ParisTech, PSL Research University
Talks:
- Introduction
- Deep Learning in 15 minutes
- Mathematical morphology: Learning simple operators
- Depthwise Morphological Layers
- Morphological Scale-Spaces
Practical Sessions:
-
Tutorial 0: Deep Learning in 15 minutes
-
Tutorial 1: Simple morphological operators using morpholayers
-
Tutorial 2: Learning morphological operators
-
Tutorial 3: Learning morphological layers in Fashion Mnist
-
Tutorial 4: Improving Max-Pooling layers using Dilations
-
Tutorial 5: Learning Additive Shift Equivariant Operators
-
Tutorial 6: Learning Scale-Equivariant Operators
