A Regression-Based Deep Learning Approach for Fish Fry Counting: Addressing Label Imbalance and Annotation Ambiguity

Melly Damara Chaniago; Lukman Zaman; Yosi Kristian

doi:10.34148/teknika.v15i2.1495

Authors

Melly Damara Chaniago Department of Information Technology, Faculty of Science and Technology, Institut Sains Teknologi Terpadu Surabaya, Surabaya, East Java, Indonesia
Lukman Zaman Department of Information Technology, Faculty of Science and Technology, Institut Sains Teknologi Terpadu Surabaya, Surabaya, East Java, Indonesia
Yosi Kristian Department of Information Technology, Faculty of Science and Technology, Institut Sains Teknologi Terpadu Surabaya, Surabaya, East Java, Indonesia

DOI:

https://doi.org/10.34148/teknika.v15i2.1495

Keywords:

Deep Learning, EfficientNetV2S, Fish Fry Counting, Annotation Refinement, Transfer Learning

Abstract

Manual counting of Nile tilapia fry is time-consuming and prone to human error, particularly under high-density conditions where visual overlap creates significant annotation ambiguity. Furthermore, real-world data collection often results in skewed label distributions. To address these challenges, this study proposes a regression-based fry counting system using EfficientNetV2S with transfer learning. Using a dataset collected from a real hatchery environment, targeted count-level balancing and rigorous annotation refinement were applied prior to model training to improve label consistency and data distribution. The novelty of this study lies in the integration of a count-level balancing strategy and a systematic annotation refinement process to address label imbalance and annotation ambiguity, two critical issues that are often overlooked in regression-based fish fry counting. The proposed model was evaluated using standard regression metrics, including Mean Absolute Error (MAE), Root Mean Square Error (RMSE), coefficient of determination (R²), and Mean Absolute Percentage Error (MAPE). Experimental results showed that the best-performing model achieved an optimized MAE of 13.25 fry, RMSE of 16.93 fry, R² of 0.985, and MAPE of 7.07%. Comparative experiments further demonstrated that the proposed EfficientNetV2S architecture outperformed EfficientNetB0, ResNet50, and MobileNetV2. These findings indicate that systematically addressing annotation ambiguity and balancing training data significantly contributes to robust fry count estimation performance in practical aquaculture applications.

Downloads

Download data is not yet available.

References

[1] P. A. Indriati and H. Hafiludin, “Manajemen Kualitas Air Pada Pembenihan Ikan Nila (Oreochromis niloticus) Di Balai Benih Ikan Teja Timur Pamekasan,” Juvenil:Jurnal Ilmiah Kelautan dan Perikanan, vol. 3, no. 2, pp. 27–31, Aug. 2022, doi: 10.21107/juvenil.v3i2.15812.

[2] J. Zhang et al., “Shrimp egg counting with fully convolutional regression network and generative adversarial network,” Aquac. Eng., vol. 94, no. June, p. 102175, 2021, doi: 10.1016/j.aquaeng.2021.102175.

[3] T. H. Khai, S. N. H. S. Abdullah, M. K. Hasan, and A. Tarmizi, “Underwater Fish Detection and Counting Using Mask Regional Convolutional Neural Network,” Water (Switzerland), vol. 14, no. 2, pp. 1–23, 2022, doi: 10.3390/w14020222.

[4] C. S. Costa et al., “Counting tilapia larvae using images captured by smartphones,” Smart Agricultural Technology, vol. 4, Aug. 2023, doi: 10.1016/j.atech.2022.100160.

[5] D. O. de Carvalho, L. F. T. Monteiro, F. M. Bazilio, G. T. H. Higa, and H. Pistori, “Livestock Fish Larvae Counting using DETR and YOLO based Deep Networks,” Aug. 2024, [Online]. Available: http://arxiv.org/abs/2408.05032

[6] A. Mumuni and F. Mumuni, “Data augmentation with automated machine learning: approaches and performance comparison with classical data augmentation methods,” Mar. 2025, doi: 10.1007/s10115-025-02349-x.

[7] A. S. Dyah Kurniasari, S. Fauziati, R. Hartanto, and A. Akmaluddin, “Nannofossil Image Enhancement for Improved Detection Using Modified Contrast Stretching Method and YOLO-v8,” in 2024 16th International Conference on Information Technology and Electrical Engineering (ICITEE), IEEE, Oct. 2024, pp. 177–182. doi: 10.1109/ICITEE62483.2024.10808703.

[8] N. E. Khalifa, M. Loey, and S. Mirjalili, “A comprehensive survey of recent trends in deep learning for digital images augmentation,” Artif. Intell. Rev., vol. 55, no. 3, pp. 2351–2377, Mar. 2022, doi: 10.1007/s10462-021-10066-4.

[9] Z. Li, X. Zhu, H. Yu, Q. Zhang, and Y. Jiang, “Edge-Aware Monocular Dense Depth Estimation with Morphology,” in 2020 25th International Conference on Pattern Recognition (ICPR), IEEE, Jan. 2021, pp. 2935–2942. doi: 10.1109/ICPR48806.2021.9412578.

[10] M. Tan and Q. V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” in Proceedings of the 36th International Conference on Machine Learning, Proceedings of Machine Learning Research, vol. 97, pp. 6105–6114, 2019”

[11] R. Ali Dihin, “Breast Cancer Detection and Diagnosis Using Gabor Features and EfficientNetV2 Model,” Journal of Al-Qadisiyah for Computer Science and Mathematics, vol. 16, no. 4, Dec. 2024, doi: 10.29304/jqcsm.2024.16.41791.

[12] P. Ye et al., “Partial Fine-Tuning: A Successor to Full Fine-Tuning for Vision Transformers,” arXiv preprint arXiv:2312.15681, 2023. [Online]. Available: http://arxiv.org/abs/2312.15681

[13] S. Eberle, A. Jentzen, A. Riekert, and G. Weiss, “Normalized Gradient Flow Optimization in the Training of ReLU Artificial Neural Networks,” arXiv preprint arXiv:2207.06246, 2022. [Online]. Available: http://arxiv.org/abs/2207.06246

[14] I. Al-Shourbaji et al., “A Deep Batch Normalized Convolution Approach for Improving COVID-19 Detection from Chest X-ray Images,” Pathogens, vol. 12, no. 1, Jan. 2023, doi: 10.3390/pathogens12010017.

[15] Z. Liu, Z. Xu, J. Jin, Z. Shen, and T. Darrell, “Dropout Reduces Underfitting,” in Proc. 40th Int. Conf. Mach. Learn. (ICML), Proc. Mach. Learn. Res., vol. 202, pp. 22233–22248, 2023.

[16] M. Reyad, A. M. Sarhan, and M. Arafa, “A modified Adam algorithm for deep neural network optimization,” Neural Comput. Appl., vol. 35, no. 23, pp. 17095–17112, Aug. 2023, doi: 10.1007/s00521-023-08568-z.

[17] D. Chicco, M. J. Warrens, and G. Jurman, “The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation,” PeerJ Comput. Sci., vol. 7, pp. 1–24, 2021, doi: 10.7717/PEERJ-CS.623.

[18] R. Hidayat et al., “Implementasi Algoritma Random Forest Regression Untuk Memprediksi Penjualan Produksi di Supermarket,” SIMKOM, vol. 10, no. 1, pp. 101–109, Jan. 2025, doi: 10.51717/simkom.v10i1.703.

[19] F. A. Rangkuti, Khairunnisa, and S. Sundari, “Implementasi Gradient Boosting Machines Untuk Prediksi Harga Rumah Pada Jakarta Selatan,” Jurnal Kecerdasan Buatan dan Teknologi Informasi, vol. 4, no. 2, pp. 164–172, May 2025, doi: 10.69916/jkbti.v4i2.318.

[20] A. A. Ahmad and N. Aman, “Improving Colorectal Cancer Detection: Leveraging EfficientNet-B0 and Explainable AI for High-Accuracy Classification,” in 2025 IEEE 2nd International Conference on Computing, Applications and Systems, COMPAS 2025, Institute of Electrical and Electronics Engineers Inc., 2025. doi: 10.1109/COMPAS67506.2025.11381829.

[21] G. Gao, Z. Sun, G. Mu, H. Yin, and Y. Ren, “Research on marine fish classification and recognition based on an optimized ResNet50 model,” Marine and Coastal Fisheries, vol. 16, no. 6, e10317, Dec. 2024, doi: 10.1002/mcf2.10317.

[22] F. A. A. Harahap, A. N. Nafisa, E. N. D. B. Purba, and N. A. Putri, “Implementasi Algoritma Convolutional Neural Network Arsitektur Model Mobilenetv2 Dalam Klasifikasi Penyakit Tumor Otak Glioma, Pituitary Dan Meningioma,” Jurnal Teknologi Informasi, Komputer, dan Aplikasinya (JTIKA ), vol. 5, no. 1, pp. 53–61, 2023, doi: 10.29303/jtika.v5i1.234.