.png)
Enhancing the Robustness of Adaptive Class Activation Mapping (AD-CAM) Against Noisy Facial Expression Data Using Preprocessing and Adaptive Normalization
Abstract
In real-world computer vision applications, visual data is often corrupted by noise, reducing both the accuracy and interpretability of deep learning models. This study proposes an enhanced AD-CAM framework that integrates noise-aware preprocessing and adaptive normalization to improve robustness in both prediction and visual explanation. Experiments were conducted on the FER2013 facial expression dataset augmented with Gaussian, salt-and-pepper, and speckle noise. Using ResNet-50 as the backbone, the proposed method demonstrated significant gains across multiple evaluation metrics, including Robust Accuracy (RA), Drop Coherence (DC), Area Under Robustness Curve (AURC), and Signal-to-Noise Ratio (SNR). Compared to the baseline, the model achieved over 10% accuracy improvement and up to 0.16 DC reduction under noise. Qualitative visualizations showed that the improved model consistently highlighted semantically relevant facial regions, maintaining interpretability even under severe input degradation. These results support the adoption of noise-aware interpretability frameworks for more reliable and trustworthy deployment in real-world vision systems.
Keywords
Full Text:
PDFReferences
Aditi and A. Dureja, “A Review: Image Classification and Object Detection with Deep Learning,” in Proceedings of 3rd International Conference on Computing Informatics and Networks (ICCIN), 2021, pp. 69–91. doi: 10.1007/978-981-33-4604-8_6
A. Chattopadhyay, A. Sarkar, P. Howlader, and V. Balasubramanian, “Grad-CAM++: Generalized Gradient-Based Visual Explanations for Deep Convolutional Networks,” in 2018 IEEE Winter Conf. Applications of Computer Vision (WACV), pp. 839–847. doi: 10.1109/WACV.2018.00097
X. Wang, “Deep Learning in Object Recognition, Detection, and Segmentation,” Found. Trends Signal Process., vol. 8, pp. 217–382, 2016. doi: 10.1561/2000000071
T. Goswami, “Impact of Deep Learning in Image Processing and Computer Vision,” in Emerging Technologies in Data Mining and Information Security, Springer, 2018, pp. 475–485. doi: 10.1007/978-981-10-7329-8_48
H. Wang, M. Du, F. Yang, and Z. Zhang, “Score-CAM: Improved Visual Explanations Via Score-Weighted Class Activation Mapping,” arXiv preprint, 2019. [Online]. Available: https://doi.org/10.48550/arXiv.1910.01279
A. Niaz, S. Soomro, H. Zia, and K. Choi, “Increment-CAM: Incrementally-Weighted Class Activation Maps for Better Visual Explanations,” IEEE Access, vol. 12, pp. 88829–88840, 2024. doi: 10.1109/ACCESS.2024.3413859
S. Iqbal, A. N. Qureshi, M. A. Alhussein, K. Aurangzeb, and M. S. Anwar, “AD-CAM: Enhancing Interpretability of Convolutional Neural Networks With a Lightweight Framework,” IEEE J. Biomed. Health Inform., vol. PP, 2023. doi: 10.1109/JBHI.2023.3329231
D. Omeiza, S. Speakman, C. Cintas, and K. Weldemariam, “Smooth Grad-CAM++: An Enhanced Inference Level Visualization Technique for Deep Convolutional Neural Network Models,” arXiv preprint, 2019. [Online]. Available: https://doi.org/10.48550/arXiv.1908.01224
M. Chakraborty, S. Sardar, and U. Maulik, “A Comparative Analysis of Non-gradient Methods of Class Activation Mapping,” in Trends in Intelligent Computing, Springer, 2022, pp. 187–196. doi: 10.1007/978-981-99-1472-2_16
R. Yang, Q. Yang, D. Chen, F. Wang, and Y. Qiu, “Explaining Deep Learning Models for COVID-19 Detection with Grad-CAM and Novel Use of PCA,” in 2024 IEEE Int. Instrum. Meas. Technol. Conf. (I2MTC), pp. 1–6. doi: 10.1109/I2MTC60896.2024.10560613
P. Zhang, Z. Huang, X. Luo, and P. Zhao, “Robust Learning with Adversarial Perturbations and Label Noise: A Two-Pronged Defense Approach,” in Proc. 4th ACM Int. Conf. Multimedia in Asia, 2022. doi: 10.1145/3551626.3564934
A. Liu, X. Liu, H. Yu, C. Zhang, Q. Liu, and D. Tao, “Training Robust Deep Neural Networks via Adversarial Noise Propagation,” IEEE Trans. Image Process., vol. 30, pp. 5769–5781, 2021. doi: 10.1109/TIP.2021.3082317
J. Zhao, “Analyzing the Robustness of Deep Learning Against Adversarial Examples,” in 2018 56th Annu. Allerton Conf. Commun., Control, Comput., pp. 1060–1064, 2018. doi: 10.1109/ALLERTON.2018.8636048
D. Gao, Y. Zhao, Y. Yao, Z. Zhang, B. Mao, and X. Yao, “Robust Deep Learning Models Against Semantic-Preserving Adversarial Attack,” in Proc. 2023 Int. Joint Conf. Neural Netw. (IJCNN), pp. 1–8, 2023. doi: 10.1109/IJCNN54540.2023.10191198
I. Goodfellow et al., “Challenges in Representation Learning: A Report on Three Machine Learning Contests,” in Neural Information Processing, Springer, 2013, pp. 117–124. doi: 10.1007/978-3-319-03545-6_13
Note: FER2013 dataset was originally introduced in this work.
P. Panda and K. Roy, “Implicit Adversarial Data Augmentation and Robustness with Noise-Based Learning,” Neural Netw., vol. 141, pp. 120–132, 2021. doi: 10.1016/j.neunet.2021.04.008
K. Philbrick et al., “What Does Deep Learning See? Insights From a Classifier Trained to Predict Contrast Enhancement Phase From CT Images,” AJR Am. J. Roentgenol., vol. 211, no. 6, pp. 1184–1193, 2018. doi: 10.2214/AJR.18.20331
A. K. Singh, D. Chaudhuri, M. P. Singh, and S. Chattopadhyay, “Integrative CAM: Adaptive Layer Fusion for Comprehensive Interpretation of CNNs,” arXiv, 2024. doi: 10.48550/arXiv.2412.01354
M. Bany Muhammad and M. Yeasin, “Eigen-CAM: Class Activation Map Using Principal Components,” in Proc. Int. Joint Conf. Neural Netw. (IJCNN), 2020, pp. 1–7. doi: 10.1109/IJCNN48605.2020.9206626
A. Ansari, A. Kalaniya, and S. Memon, “Behavioral Analysis of Neural Network Using Various Visualization Strategies,” Int. J. Adv. Res. Sci. Comput. Technol., vol. 5, no. 4, pp. 180–188, 2021. doi: 10.48175/IJARSCT-1118
Y. Liu et al., “Optimized Dropkey-Based Grad-CAM: Toward Accurate Image Feature Localization,” Sensors, vol. 23, 2023. doi: 10.3390/s23208351
F. Clement, J. Yang, and I. Cheng, “Feature CAM: Interpretable AI in Image Classification,” arXiv, 2024. doi: 10.48550/arXiv.2403.05658
S. Iqbal et al., “AD-CAM: Enhancing Interpretability of Convolutional Neural Networks With a Lightweight Framework,” IEEE J. Biomed. Health Inform., vol. PP, 2023. doi: 10.1109/JBHI.2023.3329231
Y. Lei et al., “LICO: Explainable Models with Language-Image Consistency,” arXiv, 2023. doi: 10.48550/arXiv.2310.09821
Y. Liu et al., “Optimized Dropkey-Based Grad-CAM: Toward Accurate Image Feature Localization,” Sensors, vol. 23, 2023. doi: 10.3390/s23208351
X. Chen, J. Zhang, C. Zhao, and L. Cheng, “Understanding the Decision-Making Process of CNN in Modulation Recognition via Iterative Channel Relevance,” Signal Image Video Process., vol. 18, pp. 8457–8468, 2024. doi: 10.1007/s11760-024-03486-6
Y. Sui, T. Chen, P. Xia, S. Wang, and B. Li, “Towards Robust Detection and Segmentation Using Vertical and Horizontal Adversarial Training,” in Proc. Int. Joint Conf. Neural Netw. (IJCNN), Padua, Italy, 2022, pp. 1–8. doi: 10.1109/IJCNN55064.2022.9892759
J. A. Goodwin, O. M. Brown, and V. Helus, “Fast Training of Deep Neural Networks Robust to Adversarial Perturbations,” in 2020 IEEE High Performance Extreme Computing Conference (HPEC), Waltham, MA, USA, 2020, pp. 1–7. doi: 10.1109/HPEC43674.2020.9286256
C.-K. Yeh et al., “On the (In)fidelity and Sensitivity of Explanations,” in NeurIPS, 2019. doi: 10.48550/arXiv.1901.09392
J. Adebayo et al., “Sanity Checks for Saliency Maps,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 31, 2018. doi: 10.48550/arXiv.1810.03292
DOI: https://doi.org/10.47738/jads.v7i1.1005
Refbacks
- There are currently no refbacks.
Journal of Applied Data Sciences
| ISSN | : | 2723-6471 (Online) |
| Publisher | : | Bright Publisher |
| Website | : | http://bright-journal.org/JADS |
| : | taqwa@amikompurwokerto.ac.id (principal contact) | |
| support@bright-journal.org (technical issues) |
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0