ILPO-NET: convolution network for the recognition of arbitrary volumetric patterns

Dmitrii Zhemchuzhnikov, Sergei Grudinin

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: Volumetric data, 3DCNN, pattern recognition, rotational invariance, SO(3) invariance, SE(3) invariance

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

TL;DR: The method proposes a novel convolutional operator invariant to local spatial pattern orientations. We incorporated this operator in 3DCNNs and tested it on multiple datasets.

Abstract:

Modern spatial data analysis is built on the effective recognition of spatial patterns and learning their hierarchy. Applications to real-world volumetric data require techniques that ensure invariance not only to shifts but also to pattern rotations. While traditional methods can readily achieve translational invariance, rotational invariance possesses multiple challenges and remains an active area of research.
Here, we present ILPO-Net (Invariant to Local Patterns Orientation Network), a novel approach to handling arbitrarily shaped patterns with the convolutional operation inherently invariant to local spatial pattern orientations. Our architecture seamlessly integrates the new convolution operator and, when benchmarked on diverse volumetric datasets such as MedMNIST and CATH, demonstrates superior performance over the baselines with significantly reduced parameter counts—up to 1000 times fewer in the case of MedMNIST. Beyond these demonstrations, ILPO-Net's rotational invariance paves the way for other applications across multiple disciplines.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 5890