Bias Mitigation in Ensemble-Based Meat Freshness Classification Using Grad-CAM

Abstract

Visual biases in deep learning models, such as focusing on packaging trays instead of meat texture, reduce the reliability of computer vision systems in food safety applications. This study proposes a Grad-CAM-guided bias mitigation framework for multiclass meat freshness classification that combines explainable AI with a lightweight hybrid ensemble design. A MiniCAM attention module is integrated into MobileNetV2 to redirect model focus toward meat-specific visual cues, and its features are fused with complementary embeddings extracted from Xception. The final decision is obtained by combining the predictions of MobileNetV2 with classical classifiers (SVM and XGBoost) using test-time augmentation and grid-optimized weighted ensembling. The proposed framework achieves 99.78% accuracy on the held-out test set and 99.66% ± 0.23 average accuracy under 5-fold cross-validation, while maintaining real-time efficiency (4.3M parameters, 16.5 MB model size, and 825.1 FPS on a single GPU), and effectively suppresses non-informative background elements (e.g., packaging trays) as confirmed by Grad-CAM visualizations. These results demonstrate that integrating explainable bias mitigation with lightweight ensemble learning enables reliable and deployable meat freshness assessment for real-world food safety inspection.

Keywords

• Bias Mitigation
• Computer Vision
• Ensemble Learning
• Food Safety
• Grad-CAM
• Meat Freshness

Grad-CAM Kullanılarak Topluluk Tabanlı Et Tazeliği Sınıflandırmasında Önyargının Azaltılması

Abstract

Derin öğrenme modellerindeki görsel önyargılar, örneğin et dokusuna odaklanmak yerine ambalaj tepsilerine odaklanmak, gıda güvenliği uygulamalarında güvenilirliği azaltır. Bu çalışmada, hibrit topluluk öğrenmesiyle et tazeliği sınıflandırması için Grad-CAM rehberli bir önyargı azaltma çerçevesi önerilmektedir. Aktivasyon haritası analizi, temel modellerin sistematik olarak bilgilendirici olmayan arka plan unsurlarına dikkat ettiğini ortaya koymuştur. Bunu düzeltmek için MobileNetV2’ye bir mini kanal dikkat modülü (MiniCAM) entegre ettik, özelliklerini Xception gömüleriyle birleştirdik ve test-zamanı artırımı (TTA) ile ızgara-optimizasyonlu ağırlıklı topluluk kullanarak SVM ve XGBoost tahminlerini birleştirdik. Bu yaklaşım, modelin odağını ilgili ipuçlarına (renk, doku ve nem) başarıyla yönlendirmiştir. Bu ipuçları, taze, yarı taze ve bozulmuş sınıfları ayırt etmede kritik öneme sahiptir. Gerçek dünya verilerinden oluşan çok sınıflı bir veri seti üzerinde değerlendirilen önerilen topluluk, daha yüksek doğruluk ve dayanıklılık göstermiştir. Bu çalışma, görsel akıl yürütmeyi alan uzmanlığıyla birleştirir ve Açıklanabilir Yapay Zeka (XAI)’nın yüksek riskli gıda kalite denetiminde kullanılmasını önerir. Böylece, kaynak kısıtlı tedarik zincirlerinde güvenilir dağıtım sağlanabilir.

Keywords

• Önyargı Azaltma
• Bilgisayarla Görme
• Topluluk Öğrenmesi
• Gıda Güvenliği
• Grad-CAM
• Et Tazeliği.

References

1. [1] Karanth, S., Feng, S., Patra, D., & Pradhan, A. K. (2023). Linking microbial contamination to food spoilage and food waste: The role of smart packaging, spoilage risk assessments, and date labeling. Frontiers in Microbiology, 14, 1198124. https://doi.org/10.3389/fmicb.2023.1198124
2. [2] Shi, Y., Wang, X., Borhan, M. S., Young, J., Newman, D., Berg, E., & Sun, X. (2021). A review on meat quality evaluation methods based on non-destructive computer vision and artificial intelligence technologies. Food Science of Animal Resources, 41(4), 563. https://doi.org/10.5851/kosfa.2021.e25
3. [3] Shanawad, V. (2025, November 3). Meat freshness image dataset. Kaggle. https://www.kaggle.com/datasets/vinayakshanawad/meat-freshness-image-dataset
4. [4] Büyükarıkan, B. (2024). ConvColor DL: Concatenated convolutional and handcrafted color features fusion for beef quality identification. Food Chemistry, 460, 140795. https://doi.org/10.1016/j.foodchem.2024.140795
5. [5] Abd Elfattah, M., Ewees, A. A., Darwish, A., & Hassanien, A. E. (2025). Detection and classification of meat freshness using an optimized deep learning method. Food Chemistry, 489, 144783. https://doi.org/10.1016/j.foodchem.2025.144783
6. [6] Hidalgo, M. M., Lima, R. C., De Nadai Fernandes, E. A., Bacchi, M. A., & Sarriés, G. A. (2025). Leveraging pre-trained computer vision models for accurate classification of meat freshness. Food Chemistry, 495, 146430. https://doi.org/10.1016/j.foodchem.2025.146430
7. [7] Elangovan, P., Dhurairajan, V., Nath, M. K., Yogarajah, P., & Condell, J. (2024). A novel approach for meat quality assessment using an ensemble of compact convolutional neural networks. Applied Sciences, 14(14), 5979. https://doi.org/10.3390/app14145979
8. [8] Zhou, C., Pi, J., Chen, X., Wang, D., & Liu, J. (2025). Identification and analysis of pork freshness quality based on improved MobileNetV3. Applied Engineering in Agriculture, 41(1), 57–66. https://doi.org/10.13031/aea.16131

9. [9] Shyamala Devi, M., Arun Pandian, J., Umanandhini, D., Sakineti, A., & Jeyaraj, R. (2024, January). Meat freshness state prediction using a novel fifteen layered deep convolutional neural network. In Proceedings of the International Conference on Data Science and Network Engineering (ICDSNE 2023) (pp. 103–116). Springer. https://doi.org/10.1007/978-981-99-6755-1_9
10. [10] Tanim, S. A., Shrestha, T. E., Tanvir, K., Kabir, M. S., Mridha, M. F. & Haq, M. K. (2024, September). Single-level fusion for enhancing meat quality classification with explainable AI. In Proceedings of the IEEE International Conference on Computing, Applications and Systems (COMPAS) (pp. 1–6). IEEE. https://doi.org/10.1109/COMPAS60761.2024.10796775
11. [11] Ren, X., Wang, Y., Huang, Y., Mustafa, M., Sun, D., Xue, F., Chen, D., Xu, L., & Wu, F. (2023). A CNN-based E-Nose using time series features for food freshness classification. IEEE Sensors Journal, 23(6), 6027–6038. https://doi.org/10.1109/JSEN.2023.3241842
12. [12] Susanti, E., Ariyana, R. Y., Cahyo, E. N., Sutanta, E., & Kumalasanti, R. A. (2023, October). Beef image classification using the Inception V3 transfer learning model. In Proceedings of the IEEE 9th Information Technology International Seminar (ITIS) (pp. 1–6). IEEE. https://doi.org/10.1109/ITIS59651.2023.10420013
13. [13] Hidalgo, M. M., Lima, R. C., De Nadai Fernandes, E. A., Bacchi, M. A., & Sarriés, G. A. (2025). Leveraging pre-trained computer vision models for accurate classification of meat freshness. Food Chemistry, 495, 146430. https://doi.org/10.1016/j.foodchem.2025.146430
14. [14] Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L. (2018, June). MobileNetV2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 4510–4520). IEEE. https://doi.org/10.1109/CVPR.2018.00474
15. [15] He, K., Zhang, X., Ren, S., & Sun, J. (2016, June). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 770–778). IEEE. https://doi.org/10.1109/CVPR.2016.90
16. [16] Jocher, G., & Qiu, J. (2024). Ultralytics YOLO11 (Version 11.0.0). GitHub. https://github.com/ultralytics/ultralytics
17. [17] Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009, June). ImageNet: A large-scale hierarchical image database. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 248–255). IEEE. https://doi.org/10.1109/CVPR.2009.5206848
18. [18] Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., & Batra, D. (2017, October). Grad-CAM: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE International Conference on Computer Vision (ICCV) (pp. 618–626). IEEE. https://doi.org/10.1109/ICCV.2017.74
19. [19] Chollet, F. (2017, July). Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 1251–1258). IEEE. https://doi.org/10.1109/CVPR.2017.195