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.
Ethical Statement
The authors declare that they comply with all ethical standards.
Supporting Institution
The authors declares that no funding was used in the study.
Thanks
The authors declares that no funding was used in the study.
References
- [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] 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] Shanawad, V. (2025, November 3). Meat freshness image dataset. Kaggle. https://www.kaggle.com/datasets/vinayakshanawad/meat-freshness-image-dataset
- [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] 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] 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] 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] 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] 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] 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] 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] 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] 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] 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] 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] Jocher, G., & Qiu, J. (2024). Ultralytics YOLO11 (Version 11.0.0). GitHub. https://github.com/ultralytics/ultralytics
- [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] 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] 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
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.
Ethical Statement
Yazarlar, tüm etik standartlara uyduklarını beyan etmektedir.
Supporting Institution
Yazarlar, çalışmada herhangi bir fon kullanılmadığını beyan etmektedir.
Thanks
Yazarlar, çalışmada herhangi bir fon kullanılmadığını beyan etmektedir.
References
- [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] 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] Shanawad, V. (2025, November 3). Meat freshness image dataset. Kaggle. https://www.kaggle.com/datasets/vinayakshanawad/meat-freshness-image-dataset
- [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] 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] 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] 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] 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] 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] 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] 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] 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] 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] 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] 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] Jocher, G., & Qiu, J. (2024). Ultralytics YOLO11 (Version 11.0.0). GitHub. https://github.com/ultralytics/ultralytics
- [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] 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] 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
Details
| Primary Language | English |
|---|---|
| Subjects | Computer Software |
| Journal Section | Research Article |
| Authors | |
| Submission Date | November 5, 2025 |
| Acceptance Date | March 10, 2026 |
| Publication Date | March 28, 2026 |
| DOI | https://doi.org/10.17694/bajece.1817907 |
| IZ | https://izlik.org/JA55LZ88PY |
| Published in Issue | Year 2026 Volume: 14 |
Aim & Scope
Balkan Journal of Electrical and Computer Engineering - (BAJECE)
(An International Peer Reviewed, indexed and Open Access Journal)
The journal publishes original papers in the extensive field of Electrical-electronics and Computer engineering. It accepts contributions which are fundamental for the development of electrical engineering and its applications, including overlaps to physics. Manuscripts on both theoretical and experimental work are welcome. Review articles and letters to the editors are also included.
Application areas include (but are not limited to):
Electrical & Electronics Engineering
Computer Engineering
Software Engineering
Biomedical Engineering
Electrical Power Engineering
Electric analysis & Mathematical models.
Electrical Control Engineering
Signal and Image Processing
Communications
Networking
Sensors, Actuators, Remote sensing, Consumer electronics, Fiber optics
Radar and Sonar systems
Artificial Intelligence and its Applications, Expert Systems
Medical imaging, Biomedical analysis and its applications
Computer vision, Pattern recognition,
Robotics, industrial automation
Renewable Energy
Author Guidelines
IMPORTANT INFORMATION FOR AUTHORS
Manuscript will be considered only on the understanding that it is not currently being submitted to other journals.
Submission of an article implies that the results described has not been published previously (except in the form of an abstract or as part of a published lecture or academic thesis), that it is not under consideration for publication elsewhere, that its publication is approved by all authors and that, if accepted, it will not be published elsewhere in the same form, in English or in any other language, without the written consent of the Publisher.
Authors are obliged to follow remarks and comments of the reviewers, as well as to follow Instructions for preparing manuscripts and technical remarks and corrections of the Editorial Board.
Papers will be published in English, and authors are obliged to submit papers free of typing errors (please use Spell Checking).
Editorial Board has right to make small lector corrections and to shorten text of the paper, which would not influence author's ideas and presentation. After computer lay-out of the paper, authors will obtain text as .pdf file for approval.
The manuscripts, not exceeding 12 pages (for review papers maximum 20 pages), including figures and tables, should be submitted in electronic form only in MS Word (.doc file) or LaTex format (.tex file), and submit to submission system.
For each author full name, personal (or first) name(s) and family (or last) name(s), and affiliation, have to be given. Family or last name(s) has to be written in capital letters. In the case of more than one author, the name of corresponding author should be indicated, with full postal address, e-mail address, and affiliation.
Each paper has to be written according to following order: title, author(s), affiliation(s), abstract, key words, body of the text with numerated sections and subsections including introduction, conclusions, acknowledgment (if necessary), nomenclature, and references. Pages must have page numbers.
Each author and coauthor have the possibility to subscribe for 10 reprints of their paper (for details, see our subscription page).
Authors note that paper can not be withdrawn at any condition once it is accepted. The Authors may not publish his/her The Author may not publish his/her contribution anywhere else without the prior written permission of the publisher unless it has been changed substantially. The Author warrants that his/her contribution is original, except for such excerpts from copyrighted works as may be included with the permission of the copyright holder and author thereof, that it contains no libellous statements, and does not infringe on any copyright, trademark, patent, statutory right, or propriety right of others. The Author signs for and accepts responsibility for releasing this material on behalf of any and all co-authors. In return for these rights:
All proprietary rights other than copyrights, such as patent rights.
The right to use all or part of this article, including tables and figures in future works of their own, provided that the proper acknowledgment is made to the Publisher as copyright holder.
The right to make copies of this article for his/her own use, but not for sale.
It is the responsibility of each author to ensure that papers submitted to the journal are written with ethical standards in mind, concerning plagiarism. Please note that all submissions are thoroughly checked for plagiarism. If an attempt at plagiarism is found in a published paper, the authors will be asked to issue a written apology to the authors of the original material. Any paper which shows obvious signs of plagiarism will be automatically rejected and its authors may be banned for duration of 10 years from publishing in Journal. The authors will receive proper notification if such a situation arises.
This paper has not been published in the same form elsewhere.
It will not be submitted anywhere else for publication prior to acceptance/rejection by BAJECE.
Ethics;
Papers must be submitted on the understanding that they have not been published elsewhere and are not currently under consideration by another journal. The submitting author is responsible for ensuring that the article’s publication has been approved by all the other coauthors. When an author discovers a significant error or inaccuracy in his/her own published work, it is the author's obligation to notify the publisher and cooperate with the editor to retract or correct the paper. It is also the authors’ responsibility to ensure that the articles emanating from a particular institution are submitted with the approval of the necessary institution. Only an acknowledgment from the editorial office officially establishes the date of receipt. Further correspondence and proofs will be sent to the author(s) before publication unless otherwise indicated. It is a condition of submission of a paper that the authors permit editing of the paper for readability.
BAJECE is committed to following the Code of Conduct and Best Practice Guidelines of COPE (Committee on Publication Ethics) . It is a duty of our editors to follow Cope Guidance for Editors and our peer-reviewers must follow COPE Ethical Guidelines for Peer Reviewers .
If you have any questions, please contact the relevant editorial office, or Balkan Journal of Electrical and Computer Engineering (BAJECE)' ethics representative: bajece@hotmail.com
Download a PDF version of the Ethics and Policies [PDF,392KB].
Reviewer Process Information
BAJECE employs a single-blind peer review process to ensure scientific quality, fairness, and transparency. In this review model, reviewers are able to see the authors’ names and affiliations, while authors do not have access to the reviewers’ identities. This approach allows reviewers to provide objective, detailed, and constructive feedback while maintaining their anonymity.
All submitted manuscripts are first evaluated by the Editorial Board for relevance, structure, and adherence to journal guidelines. Papers that meet the initial criteria are then assigned to at least two independent reviewers who are experts in the related research area. Reviewers assess manuscripts based on originality, technical accuracy, clarity, methodology, and scientific contribution.
Authors are required to revise their papers according to reviewers’ comments and suggestions within the given time frame. The final publication decision—acceptance, revision, or rejection—is made by the Editor-in-Chief after considering the reviewers’ recommendations and the scientific merit of the manuscript.
This single-blind review process ensures impartial evaluation, promotes academic integrity, and supports high-quality scientific publication standards in BAJECE.
Ethical Principles and Publication Policy
Papers must be submitted on the understanding that they have not been published elsewhere and are not currently under consideration by another journal. The submitting author is responsible for ensuring that the article’s publication has been approved by all the other coauthors. When an author discovers a significant error or inaccuracy in his/her own published work, it is the author's obligation to notify the publisher and cooperate with the editor to retract or correct the paper. It is also the authors’ responsibility to ensure that the articles emanating from a particular institution are submitted with the approval of the necessary institution. Only an acknowledgment from the editorial office officially establishes the date of receipt. Further correspondence and proofs will be sent to the author(s) before publication unless otherwise indicated. It is a condition of submission of a paper that the authors permit editing of the paper for readability.
BAJECE is committed to following the Code of Conduct and Best Practice Guidelines of COPE (Committee on Publication Ethics) . It is a duty of our editors to follow Cope Guidance for Editors and our peer-reviewers must follow COPE Ethical Guidelines for Peer Reviewers .
If you have any questions, please contact the relevant editorial office, or Balkan Journal of Electrical and Computer Engineering (BAJECE)' ethics representative: bajece@hotmail.com
Download a PDF version of the Ethics and Policies [PDF,392KB].
Reviewer Process Information
BAJECE employs a single-blind peer review process to ensure scientific quality, fairness, and transparency. In this review model, reviewers are able to see the authors’ names and affiliations, while authors do not have access to the reviewers’ identities. This approach allows reviewers to provide objective, detailed, and constructive feedback while maintaining their anonymity.
All submitted manuscripts are first evaluated by the Editorial Board for relevance, structure, and adherence to journal guidelines. Papers that meet the initial criteria are then assigned to at least two independent reviewers who are experts in the related research area. Reviewers assess manuscripts based on originality, technical accuracy, clarity, methodology, and scientific contribution.
Authors are required to revise their papers according to reviewers’ comments and suggestions within the given time frame. The final publication decision—acceptance, revision, or rejection—is made by the Editor-in-Chief after considering the reviewers’ recommendations and the scientific merit of the manuscript.
This single-blind review process ensures impartial evaluation, promotes academic integrity, and supports high-quality scientific publication standards in BAJECE.
Price Policy
Indexes
Journal Boards
All articles published by BAJECE are licensed under the Creative Commons Attribution 4.0 International License. This permits anyone to copy, redistribute, remix, transmit and adapt the work provided the original work and source is appropriately cited.