Publicación: Identifiying different carnivores’ tooth scores with deep learning algorithms: Testing the hominin shift in the balance of power
Cargando...
Fecha
2021-09-01
Autores
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática. Departamento de Inteligencia Artificial
Resumen
¿Cazadores o cazados? El poder sobre la cima de la cadena trófica durante la evolución humana sigue siendo hoy en día un tema controvertido, aunque la realidad es que los homínidos fueron ambas cosas. Sin embargo, es el momento en el que se produce el cambio de poder entre carnívoros y homínidos lo que sigue dando pie a arduos debates dentro de la comunidad científica. Muchos restos de homínidos presentan marcas de carnívoros, por lo que una buena manera de dar respuesta a la pregunta planteada es analizándolas detenidamente, ya que pueden albergar información sobre si el homínido fue fuente de alimento primaria para un depredador o si son producto de un proceso de carroñeo post-deposicional. En este trabajo comparamos marcas de diente hechas por cinco carnívoros distintos mediante diferentes algoritmos basados en redes neuronales convolucionales, primero de manera conjunta y después por parejas. Este método muestra distintos resultados según el taxón de los carnívoros, alcanzando un 92% de acierto en la clasificación de los datos de validación, por ejemplo, cuando comparamos marcas infligidas por leones (Panthera leo) y hienas manchadas (Crocuta crocuta). Además, se han probado algunos de los modelos para identificar las marcas de carnívoro que aparecen en un resto de diáfisis de fémur de un homínido que data sobre los 500.000 años. Los resultados sugieren que las marcas de diente fueron probablemente infligidas por carroñeo post-deposicional, en lugar que por depredación.
Were hominins hunted or hunters? The power over the predation hegemony during human evolution is still nowadays controversial. Nevertheless, the truth is they were both at some point, but it is the shift in the balance of power what remains arguable. One of the ways of answering that question is studying carnivore modification on hominin bones. Many hominin remains present bone surface modifications (BSM) made by carnivores which could result from predators feeding on them primarily or from scavengers (post-depositional). In the present work we try using several computer vision models based on convolutional neural networks to compare five different types of carnivores, firstly jointly and then pairwise. This method shows different outcomes according to specific carnivore taxa, evidencing great results when comparing tooth scores made by lions (Panthera leo) and spotted hyenas (Crocuta crocuta), for instance, with an accuracy of 92% in the testing set. It also proves the huge potential of deep learning algorithms for correct classification of BSM and their implications. Furthermore, we apply some of the models to the tooth mark observed on a femoral shaft from a hominin dated around 500 ka. The results suggest that the carnivore modification may have more likely resulted from post-depositional scavenging instead of predation.
Were hominins hunted or hunters? The power over the predation hegemony during human evolution is still nowadays controversial. Nevertheless, the truth is they were both at some point, but it is the shift in the balance of power what remains arguable. One of the ways of answering that question is studying carnivore modification on hominin bones. Many hominin remains present bone surface modifications (BSM) made by carnivores which could result from predators feeding on them primarily or from scavengers (post-depositional). In the present work we try using several computer vision models based on convolutional neural networks to compare five different types of carnivores, firstly jointly and then pairwise. This method shows different outcomes according to specific carnivore taxa, evidencing great results when comparing tooth scores made by lions (Panthera leo) and spotted hyenas (Crocuta crocuta), for instance, with an accuracy of 92% in the testing set. It also proves the huge potential of deep learning algorithms for correct classification of BSM and their implications. Furthermore, we apply some of the models to the tooth mark observed on a femoral shaft from a hominin dated around 500 ka. The results suggest that the carnivore modification may have more likely resulted from post-depositional scavenging instead of predation.
Descripción
Categorías UNESCO
Palabras clave
Tafonomía, marcas de diente, Redes Neuronales Convolucionales, carnívoro, Aprendizaje Profundo (Deep Learning)
Citación
Centro
Facultades y escuelas::E.T.S. de Ingeniería Informática
Departamento
Inteligencia Artificial