Automated Wildlife Detection System Using YOLOv12 and ESP32-S3 Microcontroller
Article Sidebar
Main Article Content
Abstract
This research aimed to develop a wildlife detection and classification system using deep learning techniques. The model employed is YOLOv12, trained on a dataset of 12,500 wildlife images collected from Kaggle, Google, Pixabay and actual field photographs. The study involved performance comparison of 66 degree and 160 degree wide-angle lenses used in combination with infrared (HC-SR501) and microwave (RCWL-0516) motion sensors. In addition, a real-time alert system was also developed using the ESP32-S3 microcontroller to transmit captured images via Wi-Fi to a FastAPI server using the HTTPS protocol and to send alerts through a Telegram Bot. The target group of interest in detection of this research was large wildlife species considered high-risk to human-wildlife conflict including elephants, tigers, bears, and gaurs. Model performance was evaluated using standard metrics including mean Average Precision (mAP), Precision, and Recall. The results revealed that the fine-tuned YOLOv12 model for wildlife detection had a mAP@0.5 of 0.908, with the highest accuracy in classifying tigers (mAP = 0.947), followed by elephants (mAP = 0.925), gaurs (mAP = 0.901), and bears (mAP = 0.859). The camera with 66-degree lens achieved an average detection accuracy of 82.28%, higher than that with 160-degree lens (72.72%) because the image from the wide-angle lens reduced the size of the animals at a distance in the image. The developed system successfully delivered real-time notifications with high reliability. The HC-SR501 sensor was suitable for close-range detection with 84.43% detection accuracy at 4 meters for elephants and 80.46% detection accuracy at 6 meters for gaurs. On the other hand, the RCWL-0516 sensor was more accurate at long range and in environments with obstacles with 69.43% detection accuracy at 8 meters for elephants and 65.17% detection accuracy at 10 meters for bears. This developed system demonstrates the potential for cost-effective and accurate wildlife monitoring and alerting in real-world applications.
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
บทความที่ได้รับการตีพิมพ์เป็นลิขสิทธิ์ของ วารสารวิทยาศาสตร์และเทคโนโลยี มหาวิทยาลัยอุบลราชธานี
ข้อความที่ปรากฏในบทความแต่ละเรื่องในวารสารวิชาการเล่มนี้เป็นความคิดเห็นส่วนตัวของผู้เขียนแต่ละท่านไม่เกี่ยวข้องกับมหาวิทยาลัยอุบลราชธานี และคณาจารย์ท่านอื่นๆในมหาวิทยาลัยฯ แต่อย่างใด ความรับผิดชอบองค์ประกอบทั้งหมดของบทความแต่ละเรื่องเป็นของผู้เขียนแต่ละท่าน หากมีความผิดพลาดใดๆ ผู้เขียนแต่ละท่านจะรับผิดชอบบทความของตนเองแต่ผู้เดียว
References
Yan, X., Shen, B. and Li, H. 2023. Small object detection method for UAVs aerial image based on YOLOv5s. In: Proceedings of the 2023 IEEE 6th International Conference on Electronic Information and Communication Technology, July 21-24, 2023. Qingdao, China.
Sahay, A., Singh, K.V. and Ponsam, G. 2023. Multi-object detection and tracking using machine learning. In: Proceedings of the 2023 International Conference on Computer Communication and Informatics, 23-25 January 2023. Coimbatore, India.
Leonid, T.T. and et al. 2023. Human wildlife conflict mitigation using YOLO algorithm. In: Proceedings of the 2023 Eighth International Conference on Science Technology Engineering and Mathematics, 6-7 April 2023. Chennai, India.
Cardellicchio, A. and et al. 2023. Tomato detection in challenging scenarios using YOLO-based single stage detectors. In: Proceedings of the 2023 IEEE International Workshop on Metrology for Agriculture and Forestry, 6-8 November 2023. Pisa, Italy.
Chen, X. and Zhai, Y. 2023. A multi-objective traffic flow detection system based on an improved YOLOv4 algorithm. In: Proceedings of the 2023 IEEE 2nd International Conference on Electrical Engineering, Big Data and Algorithms, 24-26 February 2023. Changchun, China.
Ono, S., Nishiyama, Y. and Sezaki, K. 2022. Detecting face-mask wearing status using motion sensors in commercially available smartwatches. In: Proceedings of the 2022 IEEE International Conference on E-health Networking, Application & Services, 17-19 October 2022. Genoa, Italy.
Kumar, V.M., Ajina, A. and Deepak, D.J. 2023. Improving smart home safety with face recognition using machine learning. In: Proceedings of the 2023 International Conference on Advances in Electronics, Communication, Computing and Intelligent Information Systems, 19-21 April 2023. Bangalore, India.
Elaoua, A., Nadour, M. and Cherroun, L. 2023. Real-time people counting system using YOLOv8 object detection. In: Proceedings of the 2023 2nd International Conference on Electronics, Energy and Measurement, 28-29 November 2023. Medea, Algeria.
Nale, P. and Gite, S. 2023. Real-time weapons detection system using computer vision. In: Proceedings of the 2023 Third International Conference on Smart Technologies, Communication and Robotics, 9-10 December 2023. Sathyamangalam, India.
