UnionCAM: enhancing CNN interpretability through denoising, weighted fusion, and selective high-quality class activation mapping
Deep convolutional neural networks (CNNs) have achieved remarkable success in various computer vision tasks. However, the lack of interpretability in these models has raised concerns and hindered their widespread adoption in critical domains. Generating activation maps that highlight the regions con...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2024-11-01
|
| Series: | Frontiers in Neurorobotics |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fnbot.2024.1490198/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846168206396358656 |
|---|---|
| author | Hao Hu Hao Hu Rui Wang Hao Lin Huai Yu |
| author_facet | Hao Hu Hao Hu Rui Wang Hao Lin Huai Yu |
| author_sort | Hao Hu |
| collection | DOAJ |
| description | Deep convolutional neural networks (CNNs) have achieved remarkable success in various computer vision tasks. However, the lack of interpretability in these models has raised concerns and hindered their widespread adoption in critical domains. Generating activation maps that highlight the regions contributing to the CNN's decision has emerged as a popular approach to visualize and interpret these models. Nevertheless, existing methods often produce activation maps contaminated with irrelevant background noise or incomplete object activation, limiting their effectiveness in providing meaningful explanations. To address this challenge, we propose Union Class Activation Mapping (UnionCAM), an innovative visual interpretation framework that generates high-quality class activation maps (CAMs) through a novel three-step approach. UnionCAM introduces a weighted fusion strategy that adaptively combines multiple CAMs to create more informative and comprehensive activation maps. First, the denoising module removes background noise from CAMs by using adaptive thresholding. Subsequently, the union module fuses the denoised CAMs with region-based CAMs using a weighted combination scheme to obtain more comprehensive and informative maps, which we refer to as fused CAMs. Lastly, the activation map selection module automatically selects the optimal CAM that offers the best interpretation from the pool of fused CAMs. Extensive experiments on ILSVRC2012 and VOC2007 datasets demonstrate UnionCAM's superior performance over state-of-the-art methods. It effectively suppresses background noise, captures complete object regions, and provides intuitive visual explanations. UnionCAM achieves significant improvements in insertion and deletion scores, outperforming the best baseline. UnionCAM makes notable contributions by introducing a novel denoising strategy, adaptive fusion of CAMs, and an automatic selection mechanism. It bridges the gap between CNN performance and interpretability, providing a valuable tool for understanding and trusting CNN-based systems. UnionCAM has the potential to foster responsible deployment of CNNs in real-world applications. |
| format | Article |
| id | doaj-art-c2b7b5e7b3ea4249ae58fbb7ee9b920f |
| institution | Kabale University |
| issn | 1662-5218 |
| language | English |
| publishDate | 2024-11-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Neurorobotics |
| spelling | doaj-art-c2b7b5e7b3ea4249ae58fbb7ee9b920f2024-11-14T06:21:27ZengFrontiers Media S.A.Frontiers in Neurorobotics1662-52182024-11-011810.3389/fnbot.2024.14901981490198UnionCAM: enhancing CNN interpretability through denoising, weighted fusion, and selective high-quality class activation mappingHao Hu0Hao Hu1Rui Wang2Hao Lin3Huai Yu4The Institute of Computing, China Academy of Railway Sciences Corporation Ltd, Beijing, ChinaThe Center of National Railway Intelligent Transportation System Engineering and Technology, Beijing, ChinaThe Institute of Computing, China Academy of Railway Sciences Corporation Ltd, Beijing, ChinaXi'an Jiaotong University, Xi'an, ChinaSignal and Communication Research Institute, China Academy of Railway Sciences Corporation Ltd, Beijing, ChinaDeep convolutional neural networks (CNNs) have achieved remarkable success in various computer vision tasks. However, the lack of interpretability in these models has raised concerns and hindered their widespread adoption in critical domains. Generating activation maps that highlight the regions contributing to the CNN's decision has emerged as a popular approach to visualize and interpret these models. Nevertheless, existing methods often produce activation maps contaminated with irrelevant background noise or incomplete object activation, limiting their effectiveness in providing meaningful explanations. To address this challenge, we propose Union Class Activation Mapping (UnionCAM), an innovative visual interpretation framework that generates high-quality class activation maps (CAMs) through a novel three-step approach. UnionCAM introduces a weighted fusion strategy that adaptively combines multiple CAMs to create more informative and comprehensive activation maps. First, the denoising module removes background noise from CAMs by using adaptive thresholding. Subsequently, the union module fuses the denoised CAMs with region-based CAMs using a weighted combination scheme to obtain more comprehensive and informative maps, which we refer to as fused CAMs. Lastly, the activation map selection module automatically selects the optimal CAM that offers the best interpretation from the pool of fused CAMs. Extensive experiments on ILSVRC2012 and VOC2007 datasets demonstrate UnionCAM's superior performance over state-of-the-art methods. It effectively suppresses background noise, captures complete object regions, and provides intuitive visual explanations. UnionCAM achieves significant improvements in insertion and deletion scores, outperforming the best baseline. UnionCAM makes notable contributions by introducing a novel denoising strategy, adaptive fusion of CAMs, and an automatic selection mechanism. It bridges the gap between CNN performance and interpretability, providing a valuable tool for understanding and trusting CNN-based systems. UnionCAM has the potential to foster responsible deployment of CNNs in real-world applications.https://www.frontiersin.org/articles/10.3389/fnbot.2024.1490198/fullvisual interpretationclass activation mapCNNUnion Class Activation Mappingdenoised CAMsregion-based CAMs |
| spellingShingle | Hao Hu Hao Hu Rui Wang Hao Lin Huai Yu UnionCAM: enhancing CNN interpretability through denoising, weighted fusion, and selective high-quality class activation mapping Frontiers in Neurorobotics visual interpretation class activation map CNN Union Class Activation Mapping denoised CAMs region-based CAMs |
| title | UnionCAM: enhancing CNN interpretability through denoising, weighted fusion, and selective high-quality class activation mapping |
| title_full | UnionCAM: enhancing CNN interpretability through denoising, weighted fusion, and selective high-quality class activation mapping |
| title_fullStr | UnionCAM: enhancing CNN interpretability through denoising, weighted fusion, and selective high-quality class activation mapping |
| title_full_unstemmed | UnionCAM: enhancing CNN interpretability through denoising, weighted fusion, and selective high-quality class activation mapping |
| title_short | UnionCAM: enhancing CNN interpretability through denoising, weighted fusion, and selective high-quality class activation mapping |
| title_sort | unioncam enhancing cnn interpretability through denoising weighted fusion and selective high quality class activation mapping |
| topic | visual interpretation class activation map CNN Union Class Activation Mapping denoised CAMs region-based CAMs |
| url | https://www.frontiersin.org/articles/10.3389/fnbot.2024.1490198/full |
| work_keys_str_mv | AT haohu unioncamenhancingcnninterpretabilitythroughdenoisingweightedfusionandselectivehighqualityclassactivationmapping AT haohu unioncamenhancingcnninterpretabilitythroughdenoisingweightedfusionandselectivehighqualityclassactivationmapping AT ruiwang unioncamenhancingcnninterpretabilitythroughdenoisingweightedfusionandselectivehighqualityclassactivationmapping AT haolin unioncamenhancingcnninterpretabilitythroughdenoisingweightedfusionandselectivehighqualityclassactivationmapping AT huaiyu unioncamenhancingcnninterpretabilitythroughdenoisingweightedfusionandselectivehighqualityclassactivationmapping |