TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Learning Steerable Filters for Rotation Equivariant CNNs

Learning Steerable Filters for Rotation Equivariant CNNs

Maurice Weiler, Fred A. Hamprecht, Martin Storath

2017-11-20CVPR 2018 6Multi-tissue Nucleus SegmentationRotated MNISTColorectal Gland Segmentation:Breast Tumour Classification
PaperPDF

Abstract

In many machine learning tasks it is desirable that a model's prediction transforms in an equivariant way under transformations of its input. Convolutional neural networks (CNNs) implement translational equivariance by construction; for other transformations, however, they are compelled to learn the proper mapping. In this work, we develop Steerable Filter CNNs (SFCNNs) which achieve joint equivariance under translations and rotations by design. The proposed architecture employs steerable filters to efficiently compute orientation dependent responses for many orientations without suffering interpolation artifacts from filter rotation. We utilize group convolutions which guarantee an equivariant mapping. In addition, we generalize He's weight initialization scheme to filters which are defined as a linear combination of a system of atomic filters. Numerical experiments show a substantial enhancement of the sample complexity with a growing number of sampled filter orientations and confirm that the network generalizes learned patterns over orientations. The proposed approach achieves state-of-the-art on the rotated MNIST benchmark and on the ISBI 2012 2D EM segmentation challenge.

Results

TaskDatasetMetricValueModel
Rotated MNISTRotated MNISTTest error0.714Steerable Filter CNN
Breast Tumour ClassificationPCamAUC0.971Steerable G-CNN (C8)
Breast Tumour ClassificationPCamAUC0.969Steerable G-CNN (C8)
Breast Tumour ClassificationPCamAUC0.969Steerable G-CNN (C12)
Breast Tumour ClassificationPCamAUC0.963Steerable G-CNN (e)
Colorectal Gland Segmentation:CRAGDice0.888Steerable G-CNN (C8)
Colorectal Gland Segmentation:CRAGF1-score0.861Steerable G-CNN (C8)
Colorectal Gland Segmentation:CRAGHausdorff Distance (mm)139.5Steerable G-CNN (C8)
Colorectal Gland Segmentation:CRAGDice0.87Steerable G-CNN (C12)
Colorectal Gland Segmentation:CRAGF1-score0.855Steerable G-CNN (C12)
Colorectal Gland Segmentation:CRAGHausdorff Distance (mm)156.2Steerable G-CNN (C12)
Colorectal Gland Segmentation:CRAGDice0.869Steerable G-CNN (C12)
Colorectal Gland Segmentation:CRAGF1-score0.837Steerable G-CNN (C12)
Colorectal Gland Segmentation:CRAGHausdorff Distance (mm)164.8Steerable G-CNN (C12)
Colorectal Gland Segmentation:CRAGDice0.848Steerable G-CNN (e)
Colorectal Gland Segmentation:CRAGF1-score0.811Steerable G-CNN (e)
Colorectal Gland Segmentation:CRAGHausdorff Distance (mm)175.9Steerable G-CNN (e)
Multi-tissue Nucleus SegmentationKumarDice0.82Steerable G-CNN (C12)
Multi-tissue Nucleus SegmentationKumarHausdorff Distance (mm)55.8Steerable G-CNN (C12)
Multi-tissue Nucleus SegmentationKumarDice0.818Steerable G-CNN (C12)
Multi-tissue Nucleus SegmentationKumarHausdorff Distance (mm)54.3Steerable G-CNN (C12)
Multi-tissue Nucleus SegmentationKumarDice0.809Steerable G-CNN (C4)
Multi-tissue Nucleus SegmentationKumarHausdorff Distance (mm)54.2Steerable G-CNN (C4)
Multi-tissue Nucleus SegmentationKumarDice0.791Steerable G-CNN (e)
Multi-tissue Nucleus SegmentationKumarHausdorff Distance (mm)51Steerable G-CNN (e)

Related Papers

What's Inside Your Diffusion Model? A Score-Based Riemannian Metric to Explore the Data Manifold2025-05-16GDO: Gradual Domain Osmosis2025-01-31On the Ability of Deep Networks to Learn Symmetries from Data: A Neural Kernel Theory2024-12-16Unsupervised discovery of the shared and private geometry in multi-view data2024-08-22ContextFlow++: Generalist-Specialist Flow-based Generative Models with Mixed-Variable Context Encoding2024-06-02Progressive Conservative Adaptation for Evolving Target Domains2024-02-07Attention-Map Augmentation for Hypercomplex Breast Cancer Classification2023-10-11Virchow: A Million-Slide Digital Pathology Foundation Model2023-09-14