Deep learning and taphonomy: high accuracy in the classification of cut marks made on fleshed and defleshed bones using convolutional neural networks

Abstract

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.

Artificial intelligence provides greater accuracy in the classification of modern and ancient bone surface modifications

“…CNN models are able to discriminate images on the basis of extent of microscopic features such as flaking, microstriations and shoulder effects. A comparative analysis of cut marks made with the same tools on fleshed and defleshed bones showed similar rates of correct classification when tested through CNN models that were even simpler than those used here 30 . The resolution of these deep learning methods is such that they are capable of even differentiating cut mark modifications in one-minute sequences of exposure to trampling 32 .…”

“…Carcasses were butchered in a period less than a week from acquisition for all the cut-mark sample. Important differences in the micro-features of cut marks were previously discovered when using experimental cut marks imparted on fleshed versus defleshed bones 30 . Thus, fleshed bones are a more reliable proxy when aiming to reproduce cut-mark morphologies produced during butchery and bulk defleshing.…”

“…The architecture represents a variant of the VGG16 block and repeated layer structure. As described in 30 , 32 , the model consists of a series of three blocks, each of them containing 3 × 3 kernel double layers of 32, 64 and 128 neurons respectively. In between each block, there are max-pooling (2 × 2 kernel) layers.…”

“…Computer vision accuracy increases when the targeted object can be understood within its context. This reported high accuracy of DL models has led to test the method´s ability to identify categories that are structurally almost identical, by using samples of cut marks made with the same implements on fleshed and defleshed bones 30 . The DL algorithm classified successfully up to 95% of these structurally challenging BSM.…”

Multi Region-Based Feature Connected Layer (RB-FCL) of deep learning models for bone age assessment

“…A few scientists use traditional and deep learning to estimate bone age from x-ray images. The utilization of regression to identify bone age has been utilized by a few analysts [23][24][25]. Furthermore, the utilization of random forest [26], K-NN [27], SVM [28][29][30], ANN [24,31,32], and Fuzzy Neural system [33] has been done by a few authors.…”