Abstract
Group equivariant convolutional neural networks (G-CNNs) have been successfully applied in geometric deep learning. The recently introduced framework of PDE-based G-CNNs (PDE-G-CNNs) generalizes G-CNNs while simultaneously reducing network complexity and increasing performance. In PDE-G-CNNs the usual building blocks of neural networks are replaced with solvers for evolution PDEs, these PDEs being convection, diffusion, dilation, and erosion. We investigate three geometric adaptations of PDE-G-CNNs:
-
We generalize the theory in [2] to a family of Lie groups in between roto-translation group SE(2) and Heisenberg group H(3). This geometric adaptation enables transferring training orientation score processing on SE(2) to training processing of velocity scores, shearlet transforms, or frequency scores on H(3).
-
We theoretically prove that the trainable lifting layer in a PDE-G-CNN is interchangeable with a single fixed untrained lifting coupled with multiple trainable convections. We experimentally validate this theoretical insight and report identical performance. This fixing of the lifting layer makes PDE-G-CNNs more interpretable as they now solely train association fields from neurogeometry.
-
We include curvature adaptation in PDE-G-CNNs. This curvature adaptation is beneficial within the convection part of PDE-G-CNNs as we show experimentally.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
August, J., Zucker, S.: Sketches with curvature: the curve indicator random field and Markov processes. IEEE PAMI 25(4), 387–400 (2003)
Bellaard, G., Bon, D., Pai, G., Smets, B., Duits, R.: Analysis of (sub-)Riemannian PDE-G-CNNs (2022). https://doi.org/10.48550/ARXIV.2210.00935
Cervantes-Sanchez, F., Cruz-Aceves, I., Hernandez-Aguirre, A., Hernandez-Gonzalez, M.A., Solorio-Meza, S.E.: Automatic segmentation of coronary arteries in x-ray angiograms using multiscale analysis and artificial neural networks. Appl. Sci. 9(24), 5507 (2019)
Cocci, G., Barbieri, D., Sarti, A.: Spatiotemporal receptive fields of cells in V1 are optimally shaped for stimulus velocity estimation. JOSA A 29(1), 130–138 (2012)
Cohen, T.S., Welling, M.: Group equivariant convolutional networks. In: Proceedings of the 33rd International Conference on Machine Learning, vol. 48, pp. 1–12 (2016)
Duits, R.: Perceptual organization in image analysis: a mathematical approach based on scale, orientation and curvature. Ph.D. thesis, TU/e (2005)
Duits, R., Franken, E.M.: Left-invariant parabolic evolution equations on \({SE}(2)\) and contour enhancement via invertible orientation scores, part I: Linear left-invariant diffusion equations on \({SE}(2)\). QAM-AMS 68, 255–292 (2010)
Duits, R., Führ, H., Janssen, B., Bruurmijn, M., Florack, L., van Assen, H.: Evolution equations on Gabor transforms & applications. ACHA 35(3), 483–526 (2013)
Evans, L.C.: Partial differential equations, vol. 19. AMS (2010)
Franken, E.M., Duits, R.: Crossing-preserving coherence-enhancing diffusion on invertible orientation scores. IJCV 85(3), 253–278 (2009)
Labate, D., Lim, W.Q., Kutyniok, G., Weiss, G.: Sparse multidimensional representation using shearlets. In: Papadakis, M., Laine, A.F., Unser, M.A. (eds.) Wavelets XI, vol. 5914, p. 59140U. SPIE (2005)
Liu, L., Wang, M., Zhou, S., Shu, M., Cohen, L.D., Chen, D.: Curvilinear structure tracking based on dynamic curvature-penalized geodesics. Pattern Recogn. 134, 109079 (2023)
Moiseev, I., Sachkov, Y.L.: Maxwell strata in sub-Riemannian problem on the group of motions of a plane. ESAIM Control Optim. Calc. Var. 16(2), 380–399 (2010)
Schmidt, M., Weickert, J.: Morphological counterparts of linear shift-invariant scale-spaces. JMIV 56(2), 352–366 (2016)
Smets, B.M.N., Portegies, J.W., Bekkers, E.J., Duits, R.: PDE-based group equivariant convolutional neural networks. JMIV 65(1), 209–239 (2022)
Smets, B.M.N., Portegies, J.W., St-Onge, E., Duits, R.: Total variation and mean curvature PDEs on the homogeneous space of positions and orientations. JMIV 63(2), 237–262 (2021)
Zhang, J., Dashtbozorg, B., Bekkers, E., Pluim, J.P.W., Duits, R., ter Haar Romeny, B.M.: Robust retinal vessel segmentation via locally adaptive derivative frames in orientation scores. IEEE T-MI 35(12), 2631–2644 (2016)
Acknowledgements
We gratefully acknowledge the Dutch Foundation of Science NWO for funding of VICI 2020 Exact Sciences (Duits, Geometric learning for Image Analysis, VI.C. 202-031). We thank Andrii Kompanets for a Python implementation of the www.LieAnalysis.nl package, initially developed in Mathematica, which was used to construct the cake wavelets for the fixed lifting layer experiments.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Bellaard, G., Pai, G., Bescos, J.O., Duits, R. (2023). Geometric Adaptations of PDE-G-CNNs. In: Calatroni, L., Donatelli, M., Morigi, S., Prato, M., Santacesaria, M. (eds) Scale Space and Variational Methods in Computer Vision. SSVM 2023. Lecture Notes in Computer Science, vol 14009. Springer, Cham. https://doi.org/10.1007/978-3-031-31975-4_41
Download citation
DOI: https://doi.org/10.1007/978-3-031-31975-4_41
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-31974-7
Online ISBN: 978-3-031-31975-4
eBook Packages: Computer ScienceComputer Science (R0)