Benchmark of XAI Methods for Image Classification and Extension of Optimization-Based Explanations to CLIP Models

• Rheidner A. C. S. Fernandes UFS
• Leonardo N. Matos UFS
• André B. Carvalho UFS
• Marcelo H. L. Barreto UFPE
• Flávio Arthur O. Santos UFPE

DOI: https://doi.org/10.5753/eniac.2026.30949

Resumo

Técnicas de Inteligência Artificial Explicável (XAI) são utilizadas para compreender as decisões de modelos black-box. Na classificação de imagens, métodos baseados em gradientes e otimização buscam identificar as regiões mais relevantes para as predições dos modelos. Neste trabalho, é realizado um benchmark quantitativo envolvendo Saliency Maps, Integrated Gradients, Guided Backpropagation, Grad-CAM, Guided Grad-CAM e variantes da família IPHA. Também é proposta uma extensão dos métodos de otimização para modelos multimodais baseados na arquitetura CLIP, avaliada qualitativamente de forma exploratória. Os resultados indicam vantagem para métodos por otimização que utilizam regiões contínuas, especialmente aqueles baseados em inpainting, além de demonstrar a flexibilidade desses métodos por meio de sua aplicação ao modelo CLIP.

Referências

Barreto, M. H., Oliveira, C. L., Santos, F. A., Novais, P., Matos, L. N., and Britto, A. (2025). Genetic algorithm to understand image classification. In World Conference on Information Systems and Technologies, pages 541–553. Springer.

Blank, J. and Deb, K. (2020). pymoo: Multi-objective optimization in python. IEEE Access, 8:89497–89509.

Chen, L., Li, S., Bai, Q., Yang, J., Jiang, S., and Miao, Y. (2021). Review of image classification algorithms based on convolutional neural networks. Remote Sensing, 13(22):4712.

Deb, K., Pratap, A., Agarwal, S., and Meyarivan, T. (2002). A fast and elitist multiobjective genetic algorithm: Nsga-ii. IEEE transactions on evolutionary computation, 6(2):182–197.

Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. (2009). Imagenet: A large-scale hierarchical image database. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 248–255.

Fernandes, R., Matos, L., Andre Carvalho, M. B., and Santos, F. (2026). Optimization methods for generating saliency masks in image classifiers. In IEEE World Congress on Computational Intelligence (WCCI). In press.

Hassija, V., Chamola, V., Mahapatra, A., Singal, A., Goel, D., Huang, K., Scardapane, S., Spinelli, I., Mahmud, M., and Hussain, A. (2024). Interpreting black-box models: a review on explainable artificial intelligence. Cognitive Computation, 16(1):45–74.

He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778.

Howard, J. and Gugger, S. (2020). fastai: A layered api for deep learning. Information, 11(2):108.

Iandola, F. N., Han, S., Moskewicz, M. W., Ashraf, K., Dally, W. J., and Keutzer, K. (2016). Squeezenet: Alexnet-level accuracy with 50x fewer parameters and¡ 0.5 mb model size. arXiv preprint arXiv:1602.07360.

Krause, J., Stark, M., Deng, J., and Fei-Fei, L. (2013). 3d object representations for fine-grained categorization. In Proceedings of the IEEE international conference on computer vision workshops, pages 554–561.

Lin, K.-Y., Liu, Y., Li, L., and Dou, R. (2021). A review of explainable artificial intelligence. In IFIP International Conference on Advances in Production Management Systems, pages 574–584. Springer.

Naritomy, O., Bhardwaj, S., et al. (2020). Captum: A unified and generic model interpretability library for pytorch.

Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al. (2019). Pytorch: An imperative style, high-performance deep learning library. Advances in Neural Information Processing Systems, 32.

Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al. (2021). Learning transferable visual models from natural language supervision. In International conference on machine learning, pages 8748–8763. PmLR.

Santos, F. A. O., Zanchettin, C., Silva, J. V. S., Matos, L. N., and Novais, P. (2021). A hybrid post hoc interpretability approach for deep neural networks. In International Conference on Hybrid Artificial Intelligence Systems, pages 600–610. Springer.

Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., and Batra, D. (2017). Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision, pages 618–626.

Simonyan, K., Vedaldi, A., and Zisserman, A. (2013). Deep inside convolutional networks: Visualising image classification models and saliency maps. arXiv preprint arXiv:1312.6034.

Simonyan, K. and Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.

Springenberg, J. T., Dosovitskiy, A., Brox, T., and Riedmiller, M. (2014). Striving for simplicity: The all convolutional net. arXiv preprint arXiv:1412.6806.

Sundararajan, M., Taly, A., and Yan, Q. (2017). Axiomatic attribution for deep networks. In International conference on machine learning, pages 3319–3328. PMLR.

Telea, A. (2004). An image inpainting technique based on the fast marching method. Journal of graphics tools, 9(1):23–34.

Wolf, T., Debut, L., Sanh, V., Chaumond, J., Delangue, C., Moi, A., Cistac, P., Rault, T., Louf, R., Funtowicz, M., et al. (2020). Transformers: State-of-the-art natural language processing. In Proceedings of the 2020 conference on empirical methods in natural language processing: system demonstrations, pages 38–45.