Enhancing Long-Tailed Macaque Pest Identification Using a Two-Stage YOLO-EfficientNet Pipeline

Hammami Rifqi Darmawan; Finki Dona Marleny; Ayu Ahadi Ningrum

doi:10.36456/1hrtgc63

Authors

Hammami Rifqi Darmawan Universitas Muhammadiyah Banjarmasin
Finki Dona Marleny Universitas Muhammadiyah Banjarmasin
Ayu Ahadi Ningrum Universitas Muhammadiyah Banjarmasin

DOI:

https://doi.org/10.36456/1hrtgc63

Keywords:

IoT, YOLOv8, EfficientNet, long-tailed macaque, pest repellent, smart agriculture

Abstract

Long-tailed macaque (Macaca fascicularis) infestations have reduced corn production from a potential 7–9 tons to only 4–5 tons per growing season (±40%), while conventional deterrence methods such as loud noise and shooting have proven ineffective. This study developed an Internet of Things (IoT)-based pest repellent system using audiosonic waves, an HC-SR501 sensor, and an ESP32-CAM module, incorporating YOLOv8 and EfficientNet-B0 to replace the previously used YOLOv3 Tiny model, with the aim of improving detection accuracy and distinguishing long-tailed macaques from humans. Since population density at the partner's farmland was insufficient to construct an adequate dataset, image data were collected at Kembang Island, South Kalimantan, which serves as a natural habitat supporting a large and observable population of long-tailed macaques, thereby enabling the acquisition of visually representative and diverse training data. A total of 3,083 images were annotated using Label Studio under two classes (human and long-tailed macaque), then divided at an 80:10:10 ratio without augmentation. The YOLOv8n model was trained for 50 epochs, after which validation detections were cropped into 920 images to fine-tune EfficientNet-B0 over 10 epochs. Evaluation results show that YOLOv8 achieved a precision of 0.959, recall of 0.965, mAP50 of 0.979, and mAP50-95 of 0.723, while EfficientNet-B0 reached validation accuracies of 98.4–100% with no significant overfitting. The two-stage detection-classification pipeline was successfully tested end-to-end on 121 field images, demonstrating substantial performance improvements over the prior model and contributing toward IoT and AI-based smart agriculture solutions for effective, sustainable corn field protection.

References

[1] W. Fitria, A. Nur Bambang, and J. Wasiq Hidayat, “Human and long-tailed macaque conflict in Central Java, Indonesia,” E3S Web of Conferences, vol. 202, p. 06011, Nov. 2020, doi: 10.1051/e3sconf/202020206011.

[2] K. Ilham, “Monkeys Feeding in Farm: Assessing Patterns of Crop-Feeding Long-Tailed Macaque (Macaca fascicularis) in West Sumatra, Indonesia and its Management Conflicts,” The Rufford Foundation.

[3] S. Fatimah et al., “Development of an Ultrasonic Pest Repellent System Based on ESP32 and the Internet of Things (IoT),” Justek : Jurnal Sains dan Teknologi, vol. 9, no. 1, p. 85, Mar. 2026, doi: 10.31764/justek.v9i1.37949.

[4] Md. Akkas Ali, R. Kumar Dhanaraj, and S. Kadry, “AI-enabled IoT-based pest prevention and controlling system using sound analytics in large agricultural field,” Comput. Electron. Agric., vol. 220, p. 108844, May 2024, doi: 10.1016/j.compag.2024.108844.

[5] J. C. De Dios Garcia et al., “Comparative study of Convolutional Neural Networks performance and efficiency with YOLOv8 models applied for pest detection purposes in bean plants.,” International Journal of Combinatorial Optimization Problems and Informatics, vol. 16, no. 2, pp. 112–122, Mar. 2025, doi: 10.61467/2007.1558.2025.v16i2.603.

[6] A. Raza, M. K. Shaikh, O. Ahmed Siddiqui, A. Ali, and A. Khan, “Enhancing Agricultural Pest Management with YOLO V5: A Detection and Classification Approach,” UMT Artificial Intelligence Review, vol. 3, no. 2, Dec. 2023, doi: 10.32350/umt-air.32.02.

[7] Y. Wang, M. Wu, and Y. Shen, “Identifying the Growth Status of Hydroponic Lettuce Based on YOLO-EfficientNet,” Plants, vol. 13, no. 3, p. 372, Jan. 2024, doi: 10.3390/plants13030372.

[8] F. Marleny, Muhammad Fitriansyah, Sa’adah, Winda Astria Nuansa Saputri, Rudy Ansari, and Mambang, “Segmentasi Citra Keretakan Dinding Beton Menggunakan Teknik Perbandingan Evaluasi Metrik,” TEMATIK, vol. 10, no. 1, pp. 28–33, Jun. 2023, doi: 10.38204/tematik.v10i1.1261.

[9] N. Nagaraj, R. Likhitha, B. N. Indu, and R. V. Manjunath, “Automated approach for wildlife detection and tracking using YOLOv8 deep learning algorithm,” Discover Electronics, vol. 2, no. 1, p. 107, Dec. 2025, doi: 10.1007/s44291-025-00126-2.

[10] M. Tkachenko, M. Malyuk, A. Holmanyuk, and N. Liubimov, “Label Studio: Data labeling software,” GitHub.

[11] G. Jocher, Q. Jing, and A. Chaurasia, “Ultralytics YOLO,” GitHub.

[12] J. Terven, D.-M. Córdova-Esparza, and J.-A. Romero-González, “A Comprehensive Review of YOLO Architectures in Computer Vision: From YOLOv1 to YOLOv8 and YOLO-NAS,” Mach. Learn. Knowl. Extr., vol. 5, no. 4, pp. 1680–1716, Nov. 2023, doi: 10.3390/make5040083.

[13] A. L. D. Araújo et al., “Two-step pipeline for oral diseases detection and classification: a deep learning approach,” Frontiers in Oral Health, vol. 6, Oct. 2025, doi: 10.3389/froh.2025.1659323.

[14] M. Tan and Q. V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” Sep. 2020.

[15] Y. Zhu, Y. Zhao, Y. He, B. Wu, and X. Su, “YOLO-WildASM: An Object Detection Algorithm for Protected Wildlife,” Animals, vol. 15, no. 18, p. 2699, Sep. 2025, doi: 10.3390/ani15182699.