Oil-Painting Style Classification Using ResNet with Conditional Information Bottleneck Regularization

Automatic classification of oil-painting styles holds significant promise for art history, digital archiving, and forensic investigation by offering objective, scalable analysis of visual artistic attributes. In this paper, we introduce a deep conditional information bottleneck (CIB) framework, buil...

Full description

Saved in:
Bibliographic Details
Main Authors: Yaling Dang, Fei Duan, Jia Chen
Format: Article
Language:English
Published: MDPI AG 2025-06-01
Series:Entropy
Subjects:
Online Access:https://www.mdpi.com/1099-4300/27/7/677
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849418565151621120
author Yaling Dang
Fei Duan
Jia Chen
author_facet Yaling Dang
Fei Duan
Jia Chen
author_sort Yaling Dang
collection DOAJ
description Automatic classification of oil-painting styles holds significant promise for art history, digital archiving, and forensic investigation by offering objective, scalable analysis of visual artistic attributes. In this paper, we introduce a deep conditional information bottleneck (CIB) framework, built atop ResNet-50, for fine-grained style classification of oil paintings. Unlike traditional information bottleneck (IB) approaches that minimize the mutual information <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>I</mi><mo stretchy="false">(</mo><mi>X</mi><mo>;</mo><mi>Z</mi><mo stretchy="false">)</mo></mrow></semantics></math></inline-formula> between input <i>X</i> and latent representation <i>Z</i>, our CIB minimizes the conditional mutual information <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>I</mi><mo stretchy="false">(</mo><mi>X</mi><mo>;</mo><mi>Z</mi><mo>∣</mo><mi>Y</mi><mo stretchy="false">)</mo></mrow></semantics></math></inline-formula>, where <i>Y</i> denotes the painting’s style label. We implement this conditional term using a matrix-based Rényi’s entropy estimator, thereby avoiding costly variational approximations and ensuring computational efficiency. We evaluate our method on two public benchmarks: the Pandora dataset (7740 images across 12 artistic movements) and the OilPainting dataset (19,787 images across 17 styles). Our method outperforms the prevalent ResNet with a relative performance gain of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>13.1</mn><mo>%</mo></mrow></semantics></math></inline-formula> on Pandora and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>11.9</mn><mo>%</mo></mrow></semantics></math></inline-formula> on OilPainting. Beyond quantitative gains, our approach yields more disentangled latent representations that cluster semantically similar styles, facilitating interpretability.
format Article
id doaj-art-b7fae7f0fbce4704b93ec705385f33f4
institution Kabale University
issn 1099-4300
language English
publishDate 2025-06-01
publisher MDPI AG
record_format Article
series Entropy
spelling doaj-art-b7fae7f0fbce4704b93ec705385f33f42025-08-20T03:32:26ZengMDPI AGEntropy1099-43002025-06-0127767710.3390/e27070677Oil-Painting Style Classification Using ResNet with Conditional Information Bottleneck RegularizationYaling Dang0Fei Duan1Jia Chen2School of Art and Design, Shanxi University of Electronic Science and Technology, Linfen 041000, ChinaDepartment of Fine Arts and Craft Design, Yuncheng University, Yuncheng 044030, ChinaSchool of Art and Design, Shanxi University of Electronic Science and Technology, Linfen 041000, ChinaAutomatic classification of oil-painting styles holds significant promise for art history, digital archiving, and forensic investigation by offering objective, scalable analysis of visual artistic attributes. In this paper, we introduce a deep conditional information bottleneck (CIB) framework, built atop ResNet-50, for fine-grained style classification of oil paintings. Unlike traditional information bottleneck (IB) approaches that minimize the mutual information <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>I</mi><mo stretchy="false">(</mo><mi>X</mi><mo>;</mo><mi>Z</mi><mo stretchy="false">)</mo></mrow></semantics></math></inline-formula> between input <i>X</i> and latent representation <i>Z</i>, our CIB minimizes the conditional mutual information <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>I</mi><mo stretchy="false">(</mo><mi>X</mi><mo>;</mo><mi>Z</mi><mo>∣</mo><mi>Y</mi><mo stretchy="false">)</mo></mrow></semantics></math></inline-formula>, where <i>Y</i> denotes the painting’s style label. We implement this conditional term using a matrix-based Rényi’s entropy estimator, thereby avoiding costly variational approximations and ensuring computational efficiency. We evaluate our method on two public benchmarks: the Pandora dataset (7740 images across 12 artistic movements) and the OilPainting dataset (19,787 images across 17 styles). Our method outperforms the prevalent ResNet with a relative performance gain of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>13.1</mn><mo>%</mo></mrow></semantics></math></inline-formula> on Pandora and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>11.9</mn><mo>%</mo></mrow></semantics></math></inline-formula> on OilPainting. Beyond quantitative gains, our approach yields more disentangled latent representations that cluster semantically similar styles, facilitating interpretability.https://www.mdpi.com/1099-4300/27/7/677oil-painting style classificationconditional information bottleneckmatrix-based Rényi’s α-order entropy functional
spellingShingle Yaling Dang
Fei Duan
Jia Chen
Oil-Painting Style Classification Using ResNet with Conditional Information Bottleneck Regularization
Entropy
oil-painting style classification
conditional information bottleneck
matrix-based Rényi’s α-order entropy functional
title Oil-Painting Style Classification Using ResNet with Conditional Information Bottleneck Regularization
title_full Oil-Painting Style Classification Using ResNet with Conditional Information Bottleneck Regularization
title_fullStr Oil-Painting Style Classification Using ResNet with Conditional Information Bottleneck Regularization
title_full_unstemmed Oil-Painting Style Classification Using ResNet with Conditional Information Bottleneck Regularization
title_short Oil-Painting Style Classification Using ResNet with Conditional Information Bottleneck Regularization
title_sort oil painting style classification using resnet with conditional information bottleneck regularization
topic oil-painting style classification
conditional information bottleneck
matrix-based Rényi’s α-order entropy functional
url https://www.mdpi.com/1099-4300/27/7/677
work_keys_str_mv AT yalingdang oilpaintingstyleclassificationusingresnetwithconditionalinformationbottleneckregularization
AT feiduan oilpaintingstyleclassificationusingresnetwithconditionalinformationbottleneckregularization
AT jiachen oilpaintingstyleclassificationusingresnetwithconditionalinformationbottleneckregularization