Deep Learning Computer Vision Falls Short for Bone Surface Modification Analysis in Taphonomy
This research article critically evaluates the use of deep learning (DL)-based computer vision for classifying microscopic bone surface modifications (BSMs) in archaeological and palaeontological contexts. The authors replicate and rigorously test three previously published DL image datasets and algorithms, finding that the algorithms perform poorly when exposed to new data and that all three datasets suffer from quality issues. The study warns against the optimistic presentation of DL as a solution to taphonomic equifinality, urging extreme caution until larger, balanced, higher-quality datasets more representative of the fossil record become available.
Key quotes
Despite what previous research suggests, DL algorithms are shown to not perform as well when exposed to new data.
We additionally conclude that the quality of each of the three datasets is far from ideal for any type of analysis.
This raises considerable concerns on the optimistic presentation of DL as a means of overcoming taphonomic equifinality.
In light of this, extreme caution is advised until good quality, larger, balanced, datasets, that are more analogous with the fossil record, are available.
The concept of equifinality is a central issue in taphonomy, conditioning an analyst's ability to interpret the formation and functionality of palaeontological and archaeological sites.
From the article
You might also wanna read
Machine Learning Validates Unrecognized Transient Astronomical Phenomena in Historical Observatory Images
This research paper uses machine learning to validate the existence of previously unrecognized transient astronomical phenomena in historica
arxiv.org·2mo agoDiagnostic Accuracy of a Deep Learning Framework for Automated Evaluation of Root Canal Obturation Quality From Periapical Radiographs
Researchers argue anthropomorphic LLM attributes are non-unique using Age of Empires II neural network
This paper (arXiv:2605.31514) argues that anthropomorphic attributes often ascribed to large language models (LLMs)—such as morality or natu
arxiv.org·15d agoResearchers argue anthropomorphic LLM attributes are non-unique using Age of Empires II neural network
This paper (arXiv:2605.31514) argues that anthropomorphic attributes often ascribed to large language models (LLMs)—such as morality or natu
arxiv.org·15d agoDeep Learning-Based Measurement of Keratinized Gingiva Width Using Smartphone-Acquired Clinical Images
Field Testing Satellite-Based Bramble Detection for Hedgehog Habitat Mapping
Researchers conducted field testing of a machine learning model designed to identify bramble habitats from satellite data, specifically for
Interdisciplinary 3D animation workflow for scientifically accurate palaeoart of Sri Lankan Quaternary megafauna
This academic study examines the use of 3D animation workflows to create scientifically credible palaeoart of extinct Quaternary megafauna f
palaeo-electronica.org·16d ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.