A Hybrid Geometric Morphometric Deep Learning Approach for Cut and Trampling Mark Classification

Courtenay, Lloyd A.; Huguet, Rosa; González‐Aguilera, Diego; Yravedra, José

doi:10.3390/app10010150

Cited by 32 publications

(26 citation statements)

References 59 publications

(123 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Comparison of thin plate splines between the two groups confirm grazes to be wider and shorter in comparison with scratches, as originally observed by Courtenay et al (2019b). In general, both trampling mark samples remain highly superficial in comparison to 3D observations made on cut mark groove morphology (Courtenay et al, 2019a(Courtenay et al, , 2020.…”

Section: Geometric Morphometricssupporting

confidence: 74%

“…(2019b). In general, both trampling mark samples remain highly superficial in comparison to 3D observations made on cut mark groove morphology (Courtenay et al ., 2019a, 2020).…”

Section: Resultsmentioning

confidence: 99%

“…Here, we have been able to confirm multiple observations by authors regarding qualitative descriptions of naturally produced trampling marks (Fiorillo, 1984; Andrews & Cook, 1985; Behrensmeyer et al ., 1986; Olsen & Shipman, 1988; Nielsen, 1991; Domínguez‐Rodrigo et al ., 2009; Reynard, 2014; Fernández‐Jalvo & Andrews, 2016; Courtenay et al ., 2019b). In general, both trampling mark samples remain highly superficial in comparison with cut mark groove morphology (Courtenay et al ., 2019a, 2020). The data presented here are also able to confirm the observations made by Courtenay et al .…”

Section: Discussionmentioning

confidence: 99%

“…This method is adapted from linear algebra and calculus to define the optimum grouping for the unsupervised model provided (Thorndike, 1953). K ‐means clustering was performed in R v.3.5.3 (https://www.r-project.org/) using the PC scores obtained through geometric morphometrics (Courtenay et al ., 2019c, 2020), whereas the Kneedle algorithm was used to objectively detect the elbow‐joint (Satopa et al ., 2017), performed using the free statistical software Python v.3.7 (http://www.python.org).…”

Section: Methods and Samplementioning

confidence: 99%

See 3 more Smart Citations

Scratches and grazes: a detailed microscopic analysis of trampling phenomena

Courtenay

Huguet

Yravedra

2020

Journal of Microscopy

Self Cite

View full text Add to dashboard Cite

Summary Sedimentary abrasion and postdepositional damage to fossil remains are of great interest if considering the possible distortion they could produce in the archaeological and paleontological record. Since their discovery, natural agents such as trampling phenomena have been a topic of great taphonomic interest. Nevertheless, the majority of investigation into these traces has focused almost exclusively on their differentiation from other anthropic agents such as cut marks. In recent years, advances into bone surface modification analysis via geometric morphometrics have proven useful for in‐depth characterization of different taphonomic traces; including cut, tooth and percussion marks. Through this, a preliminary study of trampling marks using advanced 3D digital microscopy was able to detect differences between what have since been known as scratch and graze marks. The present study expands from this, developing a more detailed analysis of these traces. Here, we use advanced data science techniques to provide a means of understanding trampling mark variations, contributing to our knowledge of site formation processes. Our results show how scratch and graze marks are a product of progressional decay and changes in cortical hardness, providing a new means of understanding taphonomic processes. Lay Description The study of microscopic bone surface modifications in archaeology and palaeontology is of great importance, allowing for a detailed reconstruction of the formation of a site and providing a means of interpreting the fossil register. The damage that sedimentary abrasion can produce, however, is likely to distort and influence these studies, thus requiring a detailed understanding of the different traces that can be found on different materials. Here, we use advanced 3D digital microscopy and pattern recognition algorithms to analyse the different marks produced in different sedimentological contexts, also controlling for other variables such as the state of the bone when buried, the type of bone and the time exposed to these types of damages. Through this detailed microscopic analysis of these types of damages, we are able to conclude that morphological variations in trampling marks are product of the state of decay when the bones are buried.

show abstract

Section: Geometric Morphometricssupporting

confidence: 74%

“…(2019b). In general, both trampling mark samples remain highly superficial in comparison to 3D observations made on cut mark groove morphology (Courtenay et al ., 2019a, 2020).…”

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Methods and Samplementioning

confidence: 99%

See 2 more Smart Citations

Scratches and grazes: a detailed microscopic analysis of trampling phenomena

Courtenay

Huguet

Yravedra

2020

Journal of Microscopy

Self Cite

View full text Add to dashboard Cite

show abstract

“…In more recent years, tasks in pattern recognition and classification have received an increase in efficiency and precision with the implementation of Artificially Intelligent Algorithms (AIAs), reporting >90% accuracy in GM applications. Among these, the most popular AIAs for classification purposes in GM currently include Support Vector Machines (SVM) [12][13][14][15], and Artificial Neural Networks (ANN) [16][17][18][19][20]. Both algorithms present distinct advantages, especially in the processing of complex high-dimensional data.…”

Section: Introductionmentioning

confidence: 99%

Geometric Morphometric Data Augmentation using Generative Computational Learning Algorithms

Courtenay¹,

González‐Aguilera²

2020

Preprint

Self Cite

View full text Add to dashboard Cite

The fossil record is notorious for being incomplete and distorted, frequently conditioning the type of knowledge that can be extracted from it. In many cases, this often leads to issues when performing complex statistical analyses, such as classification tasks, predictive modelling, and variance analyses, such as those used in Geometric Morphometrics. Here different Generative Adversarial Network architectures are experimented with, testing the effects of sample size and domain dimensionality on model performance. For model evaluation, robust statistical methods were used. Each of the algorithms were observed to produce realistic data. Generative Adversarial Networks using different loss functions produced multidimensional synthetic data significantly equivalent to the original training data. Conditional Generative Adversarial Networks were not as successful. The methods proposed are likely to reduce the impact of sample size and bias on a number of statistical learning applications. While Generative Adversarial Networks are not the solution to all sample-size related issues, combined with other pre-processing steps these limitations may be overcome. This presents a valuable means of augmenting geometric morphometric datasets for greater predictive visualization.

show abstract

Beyond the Problem of Bone Surface Preservation in Taphonomic Studies of Early and Middle Pleistocene Open-Air Sites

Pineda

Saladié

2022

J Archaeol Method Theory

View full text Add to dashboard Cite

A commonly identified problem in open-air sites is the poor preservation of bone surfaces because of the multiple agents and processes that act on them. In these assemblages, surface modifications of anthropic origin can be scarce or null, and its activity is mainly inferred through the stone tools and evidence of anthropogenic breakage. Carnivore activity is also frequent. La Mina and El Forn (Barranc de la Boella), Isernia La Pineta, and Torralba are open-air assemblages from the Early and Middle Pleistocene that have contributed to our knowledge of the activities that Lower Paleolithic hominins developed in open spaces. These sites show poorly preserved bone surfaces, evidence of carnivore activity, and few indications of human use on the faunal remains, although stone tools recovered are unequivocal sign of a hominin presence at those sites. Here, we present a synthesis of the taphonomic conducted at these sites with the aim of describing how this kind of work can be conducted at Paleolithic open-air sites using several different proxies, considering the limitations commonly identified in assemblages with poorly preserved bone surfaces. The absence or scarcity of cut marks could be related to the poor preservation of the faunal remains. However, it is impossible to affirm that any such marks were originally present, as hominins may have performed activities not linked to animal carcasses. Anatomical profiles have been presented as a useful tool for reconstructing the paleoecological environments and for allowing inferences to be made about the levels of competition among large predators. The assemblages reflect similarities in the deposition type of the remains and the use of these open spaces by hominins at different times during the Lower Paleolithic.

show abstract

A Hybrid Geometric Morphometric Deep Learning Approach for Cut and Trampling Mark Classification

Cited by 32 publications

References 59 publications

Scratches and grazes: a detailed microscopic analysis of trampling phenomena

Scratches and grazes: a detailed microscopic analysis of trampling phenomena

Geometric Morphometric Data Augmentation using Generative Computational Learning Algorithms

Beyond the Problem of Bone Surface Preservation in Taphonomic Studies of Early and Middle Pleistocene Open-Air Sites

Contact Info

Product

Resources

About