SD-CAF-Gate: Swin-DenseNet Cross-Attention Fusion with Adaptive Gate para Classificação de Doenças Gastrointestinais em Dataset Desbalanceado
Resumo
Este artigo apresenta o SD-CAF-Gate, uma arquitetura híbrida CNN-Transformer com gate adaptativo na atenção cruzada, que ajusta dinamicamente a contribuição de cada fonte de informação. Avaliado no dataset GastroVision (7.930 imagens, 22 classes desbalanceadas) com validação cruzada 5-folds e sem aumento sintético de dados, o modelo supera o estado-da-arte em todas as métricas, com acurácia de 85,09% e precisão macro de 76,93%.Referências
Buda, M. et al. "A systematic study of class imbalance in medical imaging," IEEE Access, vol. 9, pp. 99616-99635, 2021.
Dosovitskiy, A. et al. "An image is worth 16x16 words: Transformers for image recognition at scale," ICLR, 2021.
Fahad, M. et al. "Deep insights into gastrointestinal health: A comprehensive analysis of GastroVision dataset," Biomed. Signal Process. Control, vol. 102, p. 107260, 2025.
He, K. et al. "Deep residual learning for image recognition," Proc. IEEE CVPR, pp. 770-778, 2016.
Huang, G. et al. "Densely connected convolutional networks," Proc. IEEE CVPR, pp. 4700-4708, 2017.
Jha, D. et al. "GastroVision: A multi-class endoscopy image dataset for computer aided gastrointestinal disease detection," Sci. Data, vol. 10, p. 125, 2023.
Liu, Z. et al. "Swin Transformer: Hierarchical vision transformer using shifted windows," Proc. IEEE/CVF ICCV, pp. 10012-10022, 2021.
Sivari, E. et al. "A new approach for gastrointestinal tract findings detection and classification," Diagnostics, vol. 13, no. 4, p. 720, 2023.
Tan, M. and Le, Q. "EfficientNet: Rethinking model scaling for convolutional neural networks," Proc. ICML, pp. 6105-6114, 2019.
Dosovitskiy, A. et al. "An image is worth 16x16 words: Transformers for image recognition at scale," ICLR, 2021.
Fahad, M. et al. "Deep insights into gastrointestinal health: A comprehensive analysis of GastroVision dataset," Biomed. Signal Process. Control, vol. 102, p. 107260, 2025.
He, K. et al. "Deep residual learning for image recognition," Proc. IEEE CVPR, pp. 770-778, 2016.
Huang, G. et al. "Densely connected convolutional networks," Proc. IEEE CVPR, pp. 4700-4708, 2017.
Jha, D. et al. "GastroVision: A multi-class endoscopy image dataset for computer aided gastrointestinal disease detection," Sci. Data, vol. 10, p. 125, 2023.
Liu, Z. et al. "Swin Transformer: Hierarchical vision transformer using shifted windows," Proc. IEEE/CVF ICCV, pp. 10012-10022, 2021.
Sivari, E. et al. "A new approach for gastrointestinal tract findings detection and classification," Diagnostics, vol. 13, no. 4, p. 720, 2023.
Tan, M. and Le, Q. "EfficientNet: Rethinking model scaling for convolutional neural networks," Proc. ICML, pp. 6105-6114, 2019.
Publicado
19/07/2026
Como Citar
CLARK, Mauro Antonio Guimarães; NEVES, Iulle de Macedo Guerra; MARQUES, Júlio Vitor Monteiro; CARVALHO FILHO, Antonio Oseas de; ARAÚJO, Flávio Henrique Duarte de.
SD-CAF-Gate: Swin-DenseNet Cross-Attention Fusion with Adaptive Gate para Classificação de Doenças Gastrointestinais em Dataset Desbalanceado. In: SEMINÁRIO INTEGRADO DE SOFTWARE E HARDWARE (SEMISH), 53. , 2026, Gramado/RS.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 1016-1021.
ISSN 2595-6205.
DOI: https://doi.org/10.5753/semish.2026.23390.
