Abstract
We build on the recently introduced PDE-G-CNN framework, which proposed the concept of non-linear morphological convolutions that are motivated by solving HJB-PDEs on lifted homogeneous spaces such as the homogeneous space of 2D positions and orientations isomorphic to \(G=SE(2)\). PDE-G-CNNs generalize G-CNNs and are provably equivariant to actions of the roto-translation group SE(2). Moreover, PDE-G-CNNs automate geometric image processing via orientation scores and allow for a meaningful geometric interpretation.
In this article, we show various functional properties of these networks:
-
(1.)
PDE-G-CNNs satisfy crucial geometric and algebraic symmetries: they are semiring quasilinear, equivariant, invariant under time scaling, isometric, and are solved by semiring group convolutions.
-
(2.)
PDE-G-CNNs exhibit a high degree of data efficiency: even under limited availability of training data they show a distinct gain in performance and generalize to unseen test cases from different datasets.
-
(3.)
PDE-G-CNNs are extendable to well-known convolutional architectures. We explore a UNet variant of PDE-G-CNNs which has a new equivariant U-Net structure with PDE-based morphological convolutions.
We verify the properties and show favorable results on various datasets.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Convections and dilations/diffusion do not commute on the non-commutative G, and the desirable order is first convection and then dilation/diffusion [4, App. B, Prop. 3].
References
Aguado Lopez, J., et al.: Semigroup theory in quantum mechanics. In: Escuela-Taller de Análisis Funcional, vol. VIII (2018)
Balogh, Z.M., Engulatov, A., Hunziker, L., Maasalo, O.E.: Functional inequalities and Hamilton-Jacobi equations in geodesic spaces. Potential Anal. 36(2), 317–337 (2012)
Bekkers, E.J., Lafarge, M.W., Veta, M., Eppenhof, K.A.J., Pluim, J.P.W., Duits, R.: Roto-translation covariant convolutional networks for medical image analysis. In: Frangi, A.F., Schnabel, J.A., Davatzikos, C., Alberola-López, C., Fichtinger, G. (eds.) MICCAI 2018. LNCS, vol. 11070, pp. 440–448. Springer, Cham (2018). https://doi.org/10.1007/978-3-030-00928-1_50
Bellaard, G., Bon, D.L., Pai, G., Smets, B.M., Duits, R.: Analysis of (sub-) Riemannian PDE-G-CNNs. J. Math. Imaging Vis., 1–25 (2023). https://doi.org/10.1007/s10851-023-01147-w
Bellaard, G., Pai, G., Bescos, J.O., Duits, R.: 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. LNCS, vol. 14009, pp. 538–550. Springer, Cham (2023). https://doi.org/10.1007/978-3-031-31975-4_41
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)
Citti, G., Sarti, A.: A cortical based model of perceptional completion in the roto-translation space. JMIV 24(3), 307–326 (2006)
Cohen, T.S., Weiler, M., Kicanaoglu, B., Welling, M.: Gauge equivariant convolutional networks and the icosahedral CNN. In: Chaudhuri, K., Salakhutdinov, R. (eds.) ICML, pp. 1321–1330. PMLR (2019)
Davidson, J.L., Hummer, F.: Morphology neural networks: an introduction with applications. Circ. Syst. Sig. Process. 12(2), 177–210 (1993)
Duits, R., Florack, L., de Graaf, J., et al.: On the axioms of scale space theory. JMIV 20, 267–298 (2004). https://doi.org/10.1023/B:JMIV.0000024043.96722.aa
Duits, R., Franken, E.: Left-invariant parabolic evolution equations on \({SE}(2)\) and contour enhancement via orientation scores. QAM-AMS 68, 255–331 (2010)
Duits, R., Smets, B., Bekkers, E., Portegies, J.: Equivariant deep learning via morphological and linear scale space PDEs on the space of positions and orientations. In: Elmoataz, A., Fadili, J., Quéau, Y., Rabin, J., Simon, L. (eds.) SSVM 2021. LNCS, vol. 12679, pp. 27–39. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-75549-2_3
Happ, L.: Lax-Oleinik semi-group and weak KAM solutions. https://www.mathi.uni-heidelberg.de/~gbenedetti/13_Happ_Talk.pdf
Maragos, P., Charisopoulos, V., Theodosis, E.: Tropical geometry and machine learning. Proc. IEEE 109(5), 728–755 (2021)
Petitot, J.: The neurogeometry of pinwheels as a sub-Riemannian contact structure. J. Physiol. Paris 97, 265–309 (2003)
Renesse, M.: An optimal transport view on Schrödinger’s equation, pp. 1–11. arXiv (2009). https://arxiv.org/pdf/0804.4621.pdf
Ronneberger, O., Fischer, P., Brox, T.: U-Net: convolutional networks for biomedical image segmentation. In: Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F. (eds.) MICCAI 2015. LNCS, vol. 9351, pp. 234–241. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-24574-4_28
Sangalli, M., Blusseau, S., Velasco-Forero, S., Angulo, J.: Scale equivariant neural networks with morphological scale-spaces. In: Lindblad, J., Malmberg, F., Sladoje, N. (eds.) DGMM 2021. LNCS, vol. 12708, pp. 483–495. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-76657-3_35
Schmidt, M., Weickert, J.: Morphological counterparts of linear shift-invariant scale-spaces. J. Math. Imaging Vis. 56(2), 352–366 (2016)
Sifre, L., Mallat, S.: Rotation, scaling and deformation invariant scattering for texture discrimination. In: IEEE-CVPR, pp. 1233–1240 (2013)
Smets, B., Portegies, J., Bekkers, E.J., Duits, R.: PDE-based group equivariant convolutional neural networks. JMIV 65, 1–31 (2022). https://doi.org/10.1007/s10851-022-01114-x
Staal, J., Abràmoff, M.D., Niemeijer, M., Viergever, M.A., Van Ginneken, B.: Ridge-based vessel segmentation in color images of the retina. IEEE-TMI 23(4), 501–509 (2004)
Weiler, M., Cesa, G.: General E(2)-equivariant steerable CNNs. In: Advances in Neural Information Processing Systems, pp. 14334–14345 (2019)
Yosida, K.: Functional Analysis. CM, vol. 123. Springer, Heidelberg (1995). https://doi.org/10.1007/978-3-642-61859-8
Zuiderveld, K.: Contrast limited adaptive histogram equalization. In: Graphics Gems, pp. 474–485 (1994)
Acknowledgement
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). The git repository containing the vanilla PDE-G-CNN implementations can be found at: https://gitlab.com/bsmetsjr/lietorch.
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
Pai, G., Bellaard, G., Smets, B.M.N., Duits, R. (2023). Functional Properties of PDE-Based Group Equivariant Convolutional Neural Networks. In: Nielsen, F., Barbaresco, F. (eds) Geometric Science of Information. GSI 2023. Lecture Notes in Computer Science, vol 14071. Springer, Cham. https://doi.org/10.1007/978-3-031-38271-0_7
Download citation
DOI: https://doi.org/10.1007/978-3-031-38271-0_7
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-38270-3
Online ISBN: 978-3-031-38271-0
eBook Packages: Computer ScienceComputer Science (R0)