Enhancing Classification Models With Sophisticated Counterfactual Images
In deep learning, training data, which are mainly from realistic scenarios, often carry certain biases. This causes deep learning models to learn incorrect relationships between features when using these training data. However, because these models have <italic>black boxes</italic>, thes...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Open Journal of Signal Processing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10843353/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1823859624476934144 |
|---|---|
| author | Xiang Li Ren Togo Keisuke Maeda Takahiro Ogawa Miki Haseyama |
| author_facet | Xiang Li Ren Togo Keisuke Maeda Takahiro Ogawa Miki Haseyama |
| author_sort | Xiang Li |
| collection | DOAJ |
| description | In deep learning, training data, which are mainly from realistic scenarios, often carry certain biases. This causes deep learning models to learn incorrect relationships between features when using these training data. However, because these models have <italic>black boxes</italic>, these problems cannot be solved effectively. In this paper, we aimed to 1) improve existing processes for generating language-guided counterfactual images and 2) employ counterfactual images to efficiently and directly identify model weaknesses in learning incorrect feature relationships. Furthermore, 3) we combined counterfactual images into datasets to fine-tune the model, thus correcting the model's perception of feature relationships. Through extensive experimentation, we confirmed the improvement in the quality of the generated counterfactual images, which can more effectively enhance the classification ability of various models. |
| format | Article |
| id | doaj-art-8eec1a60cbb840cbab78e99b0c4482de |
| institution | Kabale University |
| issn | 2644-1322 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Open Journal of Signal Processing |
| spelling | doaj-art-8eec1a60cbb840cbab78e99b0c4482de2025-02-11T00:01:45ZengIEEEIEEE Open Journal of Signal Processing2644-13222025-01-016899810.1109/OJSP.2025.353084310843353Enhancing Classification Models With Sophisticated Counterfactual ImagesXiang Li0https://orcid.org/0009-0002-6012-9183Ren Togo1https://orcid.org/0000-0002-4474-3995Keisuke Maeda2https://orcid.org/0000-0001-8039-3462Takahiro Ogawa3https://orcid.org/0000-0001-5332-8112Miki Haseyama4https://orcid.org/0000-0003-1496-1761Graduate School of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido, JapanFaculty of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido, JapanData-Driven Interdisciplinary Research Emergence Department, Hokkaido University, Sapporo, Hokkaido, JapanFaculty of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido, JapanFaculty of Information Science and Technology, Hokkaido University, Sapporo, Hokkaido, JapanIn deep learning, training data, which are mainly from realistic scenarios, often carry certain biases. This causes deep learning models to learn incorrect relationships between features when using these training data. However, because these models have <italic>black boxes</italic>, these problems cannot be solved effectively. In this paper, we aimed to 1) improve existing processes for generating language-guided counterfactual images and 2) employ counterfactual images to efficiently and directly identify model weaknesses in learning incorrect feature relationships. Furthermore, 3) we combined counterfactual images into datasets to fine-tune the model, thus correcting the model's perception of feature relationships. Through extensive experimentation, we confirmed the improvement in the quality of the generated counterfactual images, which can more effectively enhance the classification ability of various models.https://ieeexplore.ieee.org/document/10843353/Counterfactual explanationcounterfactual imageimage classificationpattern recognitionand data augmentation |
| spellingShingle | Xiang Li Ren Togo Keisuke Maeda Takahiro Ogawa Miki Haseyama Enhancing Classification Models With Sophisticated Counterfactual Images IEEE Open Journal of Signal Processing Counterfactual explanation counterfactual image image classification pattern recognition and data augmentation |
| title | Enhancing Classification Models With Sophisticated Counterfactual Images |
| title_full | Enhancing Classification Models With Sophisticated Counterfactual Images |
| title_fullStr | Enhancing Classification Models With Sophisticated Counterfactual Images |
| title_full_unstemmed | Enhancing Classification Models With Sophisticated Counterfactual Images |
| title_short | Enhancing Classification Models With Sophisticated Counterfactual Images |
| title_sort | enhancing classification models with sophisticated counterfactual images |
| topic | Counterfactual explanation counterfactual image image classification pattern recognition and data augmentation |
| url | https://ieeexplore.ieee.org/document/10843353/ |
| work_keys_str_mv | AT xiangli enhancingclassificationmodelswithsophisticatedcounterfactualimages AT rentogo enhancingclassificationmodelswithsophisticatedcounterfactualimages AT keisukemaeda enhancingclassificationmodelswithsophisticatedcounterfactualimages AT takahiroogawa enhancingclassificationmodelswithsophisticatedcounterfactualimages AT mikihaseyama enhancingclassificationmodelswithsophisticatedcounterfactualimages |