Redes neuronales convolucionales (YOLO) para la detección de megafauna marina y basura marina flotante a partir de imágenes aéreas

• Maria V DeLaHoz ^[1] ; Toni Monleón-Getino ^[1] ; Albert Martín ^[1] ; Odei Garcia- Garin ^[2] ; Morgana Vighi ^[1] ; Marina Costa ^[3] ; Valentina De Santis ^[3] ; Caterina Lanfredi ^[3] ; Capt. Daniele Giannelli ^[4] ; Sabina Airoldi ^[3]
  1. [1] Universitat de Barcelona

     Universitat de Barcelona

     Barcelona, España

     
  2. [2] Universitat de Girona

     Universitat de Girona

     Gerona, España

     
  3. [3] Tethys Research Institute,Milano, Italia
  4. [4] Italian Coast Guard Headquarters. Roma, Italia

  Mostrar afiliaciones +
• Localización: RISTI: Revista Ibérica de Sistemas e Tecnologias de Informação, ISSN-e 1646-9895, Nº. Extra 78, 2026, págs. 40-54
• Idioma: español
• Títulos paralelos:
  • Convolutional Neural Networks (YOLO) for detection of marine organisms and floating marine litter from aerial images
• Enlaces
  • Texto completo (pdf)
• Resumen
  • español

    Este estudio evalúa el potencial de la red neuronal convolucional YOLOv8 para detección y clasificación de megafauna y basura flotante en el mar a partir de imágenes aéreas. Se realizaron distintos ensayos con imágenes de cetáceos, obteniendo un rendimiento satisfactorio en casi todas las especies evaluadas, con probabilidades de detección correcta, valores de mAP-50 y valores de F1-score alrededor de 0.9. El mejor compromiso entre overfitting y función de pérdida a lo largo de los ensayos, se observó al ejecutar 500 iteraciones, más allá de las cuales aumentaba fuertemente dicho sobreajuste. A modo comparativo, un ensayo con distintos ítems de basura marina mostró proporciones relativamente altas de detecciones incorrectas y métricas muy fluctuantes durante la validación, ilustrando signos claros de underfitting y sugiriendo posibilidad de mejoras gracias a la flexibilidad del modelo. Se proponen los posibles alcances de YOLOv8 como herramienta de apoyo a actividades de monitoreo para conservación marina.

  • English

    This study evaluates the potential of the convolutional neural network YOLOv8 for detection and classification of megafauna and floating marine litter, from aerial images. Several runs were conducted with images of various cetacean species, achieving satisfactory performance in most of them, with probabilities of correct detection, mAP-50 values, and F1-scores around 0.9. The best trade-off between overfitting and loss function was observed when running the algorithm for 500 epochs, beyond which a sharp increase of overfitting occurred. A comparative analysis with different items of marine litter, showed high proportions of incorrect detections and highly fluctuating metrics during validation, illustrating clear signs of underfitting. Varying contrasts between object colorations and the background, may be a factor influencing the CNN’s efficiency between classes. However, the flexibility of the model leaves possibilities of improvements. Further application of the use of YOLOv8 is proposed as a tool to support monitoring activities for marine conservation.

• Referencias bibliográficas
  • Al Abbar, F., Eikelboom, J., Costelloe, B.R., Lorenzo, F., Ramon, H., Visser, F., GonzálezGarcía, L., Aarts, G. & van-Langevelde, F. (2024)....
  • Aldaya, A., Monleón, T., Flor, N. & Justel-Pizarro, J. (2023). Object Detection with Machine Learning, a YOLO review. In: https://www....
  • Alsaidi, M., Al-Jassani, M. G., Bang, C., O’Corry-Crowe, G., Watt, C., Ghazal, M., & Zhuang, H. (2024). Localization and tracking of beluga...
  • Bocanegra-Valencia, J.P., Urrelo-Huiman, L.V. & Cieza-Mostacero, S.V. (2024). Deep Learning para la detección de Cáncer de Pulmón: Una...
  • Borowicz, A., Le, H., Humphries, G., Nehls, G., H¨oschle, C.,… & Kosarev, V. (2019). Aerial-trained deep learning networks for surveying...
  • Boulent, J., Charry, B., Kennedy, M. M., Tissier, E., Fan, R., Marcoux, M., ... & GagnéTurcotte, A. (2023). Scaling whale monitoring using...
  • Garcia-Garin, O., Monleon-Getino, A., Lopez-Brosa, P., Borrell, A., Aguilar, A., BorjaRobalino, R., Cardona, L. & Vighi M. (2021). Automatic...
  • Gray, P.C., Bierlich, K.C., Mantell, S.A., Friedlaender, A.S., Goldbogen, J.A., Johnston, D.W. & Ye, H. (2019). Drones and convolutional...
  • Green, K. M., Virdee, M. K., Cubaynes, H. C., Aviles-Rivero, A. I., Fretwell, P. T., Gray, P. C., ... & Jackson, J. A. (2023). Gray whale...
  • Guirado, E., Tabik, S., Rivas, M. L., Alcaraz-Segura, D., & Herrera, F. (2019). Whale counting in satellite and aerial images with deep...
  • Jiang J. (2024). The eye of artificial intelligence-Convolutional NeuralNetworks. Proc. 2nd International Conference on Software Engineering...
  • Jocher, G. 2023. Ultralytics YOLOv8. https://github.com/ultralytics/ultralytics (accessed Nov 10, 2025).
  • Justel-Pizarro, J.T. (2022). Clasificación automática de formas biológicas mediante deep learning. Tesis Máster Interuniversitario en Estadística...
  • Khan, A., Sohail, A., Zahoora, U. & Qureshi, A.S. (2020) A survey of the recent architectures of deep convolutional neural networks. Artificial...
  • Lin, T.Y., Maire, M., Belongie, S., Hays, J., Perona, P.,… & Ramanan, D. (2014). Microsoft COCO: common objects in context. Computer vision...
  • Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. arXiv:2305.09972v2...
  • Reis, D., Kupec, J., Hong, J., & Daoudi, A. (2024). Real-Time Flying Object Detection with YOLOv8. arXiv:2305.09972v2 [cs.CV]. https://doi.org/10.48550/ arXiv.2305.09972
  • Ren, S., He, K., Girshick, R. & Sun, J. (2015). Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. in Advances...
  • Sanikommu, V., Sura, A., & Chimirala, P. (2025). Deep Learning-Based Detection and Tracking of Fin Whales Using High-Resolution Space-Borne...
  • Solawetz, J., Francesco. (2023). What is yolov8? the ultimate guide. https://blog. roboflow.com/what-is-yolov8/
  • Smith, L. N. (2018). A disciplined approach to neural network hyper-parameters: Part 1 -- learning rate, batch size, momentum, and weight...
  • Tethys Research Institute–Cetacean Sanctuary Research. (2024). Eye in the Sky, Technical activities Report, Phase I. By Costa, M., De Santis,...
  • Wang, X., Zuraidah, R., Binti, R. & Rasi, M. (2024). Consumer Demand Prediction Algorithm Based on Deep Learning in C2M Mode Supply Chain....
  • Yaseen, M., (2024). What is YOLOv8: An in-depth exploration of the internal features of the next generation object detector. ArXiv:2408.15857v1...
  • Yu, L. & Liu, A. (2023). Single-Stage Deep Learning-based Approach for Real-Time License Plate Recognition in Smart Parking System. International...
  • Zhang A., Lipton, Z. C. Li M. & Smola A. J. (2021). Dive into Deep Learning. arXiv preprint arXiv:2106.11342, 2021.