Deep learning classification applied to traffic accidents prediction

• Coll-Josifov, Richard ^[1] ; Masip-Álvarez, Albert ^[1] ; Lavèrnia-Ferrer, David ^[1]
  1. [1] Universitat Politècnica de Catalunya

     Universitat Politècnica de Catalunya

     Barcelona, España

     
• Localización: XLIII Jornadas de Automática: libro de actas: 7, 8 y 9 de septiembre de 2022, Logroño (La Rioja) / coord. por Carlos Balaguer Bernaldo de Quirós, José Manuel Andújar Márquez, Ramón Costa Castelló, C. Ocampo-Martínez, Juan Jesús Fernández Lozano, Matilde Santos Peñas, José Simó, Montserrat Gil Martínez, José Luis Calvo Rolle, Raúl Marín, Eduardo Rocón de Lima, Elisabet Estévez Estévez, Pedro Jesús Cabrera Santana, David Muñoz de la Peña Sequedo, José Luis Guzmán Sánchez, José Luis Pitarch Pérez, Óscar Reinoso García, Óscar Déniz Suárez, Emilio Jiménez Macías, Vanesa Loureiro-Vázquez, 2022, ISBN 978-84-9749-841-8, págs. 964-971
• Idioma: inglés
• Enlaces
  • Texto completo
• Resumen

  • In this paper, YOLOv4 neural networks are trained with the goal of detecting and classifying objects from a street as seen from a drone. These have been trained on the VisDrone dataset, which is firstly validated through a custom-made graphic user interface. Then, several tests regarding performance, dataset composition and contrast have been carried out on the trained models. Results are compared to those from other previously existing models in order to evaluate their performance and analyse their shortcomings. The trained models are then used to detect and classify objects in a city scenario in real-time. Finally, an algorithm is proposed to track the objects, infer their future trajectories and predict potential collisions from the expected trajectories.