Optimization of animal detection in thermal images using YOLO architecture


  • Łukasz Popek Warsaw University of Technology, Faculty of Electronics and Information Technology, Nowowiejska 15/19, 00-665 Warszawa, Poland http://orcid.org/0000-0002-9560-5939
  • Rafał Perz Warsaw University of Technology, Faculty of Power and Aeronautical Engineering, Nowowiejska 24, 00-665 Warszawa
  • Grzegorz Galiński Faculty of Electronics and Information Technology, Nowowiejska 15/19, 00-665 Warszawa, Poland
  • Artur Abratański Warsaw University of Technology, Faculty of Power and Aeronautical Engineering, Nowowiejska 24, 00-665 Warszawa


The article presents research on animal detection in thermal images using the YOLOv5 architecture. The goal of the study was to obtain a model with high performance in detecting animals in this type of images, and to see how changes in hyperparameters affect learning curves and final results. This manifested itself in testing different values of learning rate, momentum and optimizer types in relation to the model's learning performance. Two methods of tuning hyperparameters were used in the study: grid search and evolutionary algorithms. The model was trained and tested on an in-house dataset containing images with deer and wild boars. After the experiments, the trained architecture achieved the highest score for Mean Average Precision (mAP) of 83\%. These results are promising and indicate that the YOLO model can be used for automatic animal detection in various applications, such as wildlife monitoring, environmental protection or security systems.


J. Witczuk et al. “Exploring the feasibility of unmanned aerial vehicles andthermal imaging for ungulate surveys in forests - preliminary results. International Journal of Remote Sensing”. In: International Journal of Remote Sensing 39.15-16 (2018), pp. 5504–5521. doi: 10.1080/01431161.2017.1390621.

A. Vecvanags et al. “Ungulate Detection and Species Classification from CameraTrap Images Using Reti-naNet and Faster R-CNN”. In: Entropy 24.3 (2022), p. 353. doi: 10.3390/e24030353.

Mateusz Choiński et al. “A First Step Towards Automated Species Recognition from Camera Trap Images of Mammals Using AI in a European TemperateForest”. In: (2021), pp. 299–310. doi: 10.1007/978-3-030-84340-3 24.

M. Ivansic-Kos, M. Kristo, and M. Pobar. Human Detection in Thermal Imaging Using YOLO. 5th International Conference on Computer and Technology Applications, pp. 19-24. 2019. doi: 10.1145/3323933.3324076.

M. Kristo, M. Ivasic-Kos, and M. Pobar. “Thermal Object Detection in Difficult Weather Conditions Using YOLO”. In: IEEE Access PP.3 (2020), pp. 125459–125476. doi: 10.1109/ACCESS.2020.3007481.

Ippalapally R., S. H. Mudumba, and H. R. Adkay M. Nandi Vardhan. Human Detection in Thermal Imaging Using YOLO. Object Detection Using ThermalImaging, 17th India Council International Conference (INDICON), pp. 19-24, New Delhi, India. 2020. doi: 10.1145/3323933.3324076.

A. Ulhaq et al. “Automated Detection of Animals in Low-Resolution AirborneThermal Imagery”. In: Remote Sensing PP.3 (2021), pp. 125459–125476. doi:10.1109/ACCESS.2020.3007481.

Popek, L., Perz, R., and Galiński, G. “Comparison of Different Methods of Animal Detection and Recognition on Thermal Camera Images”. In: Electronics12.270 (2023), pp. 125459–125476. doi: 10.3390/electronics12020270.

J. Cilulko, P. Janiszewski, and M. et al. Bogdaszewski. “Infrared thermal imaging in studies of wild animals”. In: European Journal of Wildlife Researche 59 (2013), pp. 17–23. doi: 10.1007/s10344-012-0688-1.

L. Tan, T. Huangfu, and L. et al. Wu. “Comparison of RetinaNet, SSD, and8 YOLO v3 for real-time pill identification”. In: BMC Med Inform Decis Mak (2021). doi: 10.1186/s12911-021-01691-8.

Jocher Glen. YOLOv5 by Ultralytics (Version 7.0) [Computer software]. access: 13.06.2023. 2014. doi: 10.5281/zenodo.3908559.

Isa, I. S. et al. “Optimizing the Hyperparameter Tuning of YOLOv5 for Underwater Detection”. In: IEEE Access 10 (2022), pp. 52818–52831. doi: 10.1109/ACCESS.2022.3174583.

Kaichao You et al. “How does learning rate decay help modern neural networks?”In: arXiv preprint arXiv:1908.01878 (2019). doi: 10.48550/arXiv.1908.01878.

Bryan Lim, Stefan Zohren, and Stephen Roberts. “Enhancing time-series momentum strategies using deep neural networks”. In: The Journal of Financial Data Science (2019). doi: 10.3905/jfds.2019.1.015.14

Thomas M Breuel. “The effects of hyperparameters on SGD training of neuralnetworks”. In: arXiv preprint arXiv:1508.02788 (2015). doi: 10.48550/arXiv.1508.02788.

Imran Khan Mohd Jais, Amelia Ritahani Ismail, and Syed Qamrun Nisa. “Adam optimization algorithm for wide and deep neural network”. In: Knowledge Engineering and Data Science 2.1 (2019), pp. 41–46.

E. Bisong. “Google Colaboratory. In Building Machine Learning and Deep Learning Models on Google Cloud Platform”. In: (2019)


Q. Xu et al. “Effective Face Detector Based on YOLOv5 and Superresolution Reconstruction.” In: Computational and mathematical methods in medicine (2021) doi: 10.1155/2021/7748350.






Image Processing