Bone surface modifications are foundational to the correct identification of hominin butchery traces in the archaeological record. Until present, no analytical technique existed that could provide objectivity, high accuracy, and an estimate of probability in the identification of multiple structurally-similar and dissimilar marks. Here, we present a major methodological breakthrough that incorporates these three elements using Artificial Intelligence (AI) through computer vision techniques, based on convolutional neural networks. This method, when applied to controlled experimental marks on bones, yielded the highest rate documented to date of accurate classification (92%) of cut, tooth and trampling marks. After testing this method experimentally, it was applied to published images of some important traces purportedly indicating a very ancient hominin presence in Africa, America and Europe. The preliminary results are supportive of interpretations of ancient butchery in some places, but not in others, and suggest that new analyses of these controversial marks should be done following the protocol described here to confirm or disprove these archaeological interpretations.
Accurate identification of bone surface modifications (BSM) is crucial for the taphonomic understanding of archaeological and paleontological sites. Critical interpretations of when humans started eating meat and animal fat or when they started using stone tools, or when they occupied new continents or interacted with predatory guilds impinge on accurate identifications of BSM. Until now, interpretations of Plio-Pleistocene BSM have been contentious because of the high uncertainty in discriminating among taphonomic agents. Recently, the use of machine learning algorithms has yielded high accuracy in the identification of BSM. A branch of machine learning methods based on imaging, computer vision (CV), has opened the door to a more objective and accurate method of BSM identification. The present work has selected two extremely similar types of BSM (cut marks made on fleshed an defleshed bones) to test the immense potential of artificial intelligence methods. This CV approach not only produced the highest accuracy in the classification of these types of BSM until present (95% on complete images of BSM and 88.89% of images of only internal mark features), but it also has enabled a method for determining which inconspicuous microscopic features determine successful BSM discrimination. The potential of this method in other areas of taphonomy and paleobiology is enormous.
Humans are unique in their diet, physiology and socio-reproductive behavior compared to other primates. They are also unique in the ubiquitous adaptation to all biomes and habitats. From an evolutionary perspective, these trends seem to have started about two million years ago, coinciding with the emergence of encephalization, the reduction of the dental apparatus, the adoption of a fully terrestrial lifestyle, resulting in the emergence of the modern anatomical bauplan, the focalization of certain activities in the landscape, the use of stone tools, and the exit from Africa. It is in this period that clear taphonomic evidence of a switch in diet with respect to Pliocene hominins occurred, with the adoption of carnivory. Until now, the degree of carnivorism in early humans remained controversial. A persistent hypothesis is that hominins acquired meat irregularly (potentially as fallback food) and opportunistically through klepto-foraging. Here, we test this hypothesis and show, in contrast, that the butchery practices of early Pleistocene hominins (unveiled through systematic study of the patterning and intensity of cut marks on their prey) could not have resulted from having frequent secondary access to carcasses. We provide evidence of hominin primary access to animal resources and emphasize the role that meat played in their diets, their ecology and their anatomical evolution, ultimately resulting in the ecologically unrestricted terrestrial adaptation of our species. This has major implications to the evolution of human physiology and potentially for the evolution of the human brain.
Deep learning models are based on a combination of neural network architectures, optimization parameters and activation functions. All of them provide exponential combinations whose computational fitness is difficult to pinpoint. The intricate resemblance of the microscopic features that are found in bone surface modifications make their differentiation challenging, and determining a baseline combination of optimizers and activation functions for modeling seems necessary for computational economy. Here, we experiment with combinations of the most resolutive activation functions (relu, swish, and mish) and the most efficient optimizers (stochastic gradient descent (SGD) and Adam) for bone surface modification analysis. We show that despite a wide variability of outcomes, a baseline of relu–SGD is advised for raw bone surface modification data. For imbalanced samples, augmented datasets generated through generative adversarial networks are implemented, resulting in balanced accuracy and an inherent bias regarding mark replication. In summary, although baseline procedures are advised, these do not prevent to overcome Wolpert’s “no free lunch” theorem and extend it beyond model architectures.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.