Select Page

Research digital skills training 2021

Automated stone artefacts classification using machine learning

Dr Joshua Emmitt, Research Fellow, Social Science; Sina Masoud-Ansari, Centre for eResearch; Dr Rebecca Phillipps, Senior Lecturer, Anthropology; Stacey Middleton, Masters, Anthropology; Prof Simon Holdaway, Assoc Deputy VC Research, Research Strategy and Integrity

  1. Home
  2.  • 
  3. Project
  4.  • Automated stone artefacts classification using machine learning

Introduction

The ability to distinguish natural from human manufactured stone artifacts has a long history in archaeology. Methods intended to differentiate unmodified from modified objects have not developed significantly since those identified in the 20th century.
A variety of attribute based approaches are described but despite these, in both past and more recent studies, there is a continued reliance on the opinion of experts who are able to identify relevant attributes and therefore make determinations about the origin of these objects. This raises issues around access to such experts and consistency in the identifications these experts make.

Another set of issues concern the volume of material that needs to be identified. In many instances gravel sized rock clasts including those 

Methods

The Centre for eResearch developed the machine learning workflow for the automated classification system. This consisted of several components: training data generation, data preprocessing and model training. Training data was generated through expert labelling of approximately 6000 images of rocks and artefacts in various field and laboratory settings. The labels consisted of the object bounds, manually annotated by an archaeologist and the object classification (rock or artefact). The preprocessing step standardised the images and associated object bounding boxes to a 300×300 pixel format which was used to train a convolutional neural network (CNN) based object detection model. The model architecture was an openly available PyTorch implementation of the Faster R-CNN ResNet 50 which is designed to locate regions of interest in an image and generate a set of classification scores to rank the likelihood of a region matching a given class. Figure 1 summarises the object detection and classification process.

potentially showing anthropogenic or anthropoid modification are abundant, so separating classes of modified from unmodified material may involve large numbers of individual identifications. Our interest in differentiating stone artifacts from naturally occurring rocks relates to the abundance of rock clasts found in some archaeological sites. Stone artifacts are often the most abundant class of artifacts but their consistent identification is limited by the number of archaeologists with experience in their identification. The issues are compounded because many of the stone objects found come from commercial resource management projects that may lack the resources for detailed stone artifact identification. Here we report work intended to create a machine learning based technology for stone artifact identification as part of a solution to the lack of experts available to distinguish worked stone objects from naturally occurring clasts

 

Results

Model training took approximately 7 hours on an NVIDIA P40 GPU which was made available to us via the Nectar Research Cloud. The model was evaluated on a separate benchmark of approximately 1000 images which were selected to represent a range of stone artefact types. The benchmark images were also presented to expert archaeologists to compare against machine performance.

The trained model was able to achieve 100% accuracy on the benchmark images with a mean intersection over union (IoU) score of 0.82 where 1.0 indicates perfect alignment between predicted object bounds and the ground truth as determined by an expert. Two expert archaeologists were asked to classify the benchmark images and had classification accuracies of 99.4% and 94.8% respectively. The results suggest that automated detection of rocks and stone artefacts is possible and performs at near or above human expert performance on this task.

Figure 1: Object detection workflow. An input image 1) is preprocessed to a fixed size 2) with black pixel padding to meet the final dimensions. The model learns to identify potential object locations 3) which are used as input to a CNN 4) that learns feature combinations that distinguish one class from another. The final output is the region bounds and a classification e.g. rock or artefact 5).

 

See more case study projects

Developing a genomics-specific Data Management Plan (DMP) using the  Data Stewardship Wizard

Developing a genomics-specific Data Management Plan (DMP) using the Data Stewardship Wizard

Understanding the effects of Airbnb on land use, land value and regulation

Understanding the effects of Airbnb on land use, land value and regulation

Calibrating gravitational wave signal parameters of Extreme Mass Ratio Inspirals (EMRIs)

Calibrating gravitational wave signal parameters of Extreme Mass Ratio Inspirals (EMRIs)

Automated stone artefacts classification using machine learning

Automated stone artefacts classification using machine learning

Hands-on DNA: exploring the impact of virtual reality on teaching DNA structure and function

Hands-on DNA: exploring the impact of virtual reality on teaching DNA structure and function

Re-assess urban spaces under COVID-19 impact: sensing Auckland social ‘hotspots’ with mobile location data

Re-assess urban spaces under COVID-19 impact: sensing Auckland social ‘hotspots’ with mobile location data

Aotearoa New Zealand’s changing coastline – Resilience to Nature’s Challenges (National Science Challenge)

Aotearoa New Zealand’s changing coastline – Resilience to Nature’s Challenges (National Science Challenge)

Auckland housing and land use geo-data

Auckland housing and land use geo-data

Rapid monitoring of infrastructural health using remote sensing

Rapid monitoring of infrastructural health using remote sensing

Enhancing Spontaneous Recovery after Stroke Study (ESPRESSo)

Enhancing Spontaneous Recovery after Stroke Study (ESPRESSo)

Data analytics and visualisation for improving  public health and transport planning

Data analytics and visualisation for improving public health and transport planning

Data maturity project in High Value Nutrition (Phase 2) – National Science Challenge

Data maturity project in High Value Nutrition (Phase 2) – National Science Challenge

Supporting the airborne remote sensing mission – Rongowai

Supporting the airborne remote sensing mission – Rongowai

A collaborative extended reality tool to examine tumour evolution (Phase II)

A collaborative extended reality tool to examine tumour evolution (Phase II)

Data maturity project in High Value Nutrition, National Science Challenge

Data maturity project in High Value Nutrition, National Science Challenge

Haka on the move: sport circuits and cultural performance 

Haka on the move: sport circuits and cultural performance 

Proteins under a computational microscope: designing in-silico strategies to understand and develop molecular functionalities in Life Sciences and Engineering

Proteins under a computational microscope: designing in-silico strategies to understand and develop molecular functionalities in Life Sciences and Engineering

Remote temperature monitoring to reduce the spread of COVID-19

Remote temperature monitoring to reduce the spread of COVID-19

COVID-19 exponential growth visualisation

COVID-19 exponential growth visualisation

Developing virtual capabilities for the Science Payload Operations Centre

Developing virtual capabilities for the Science Payload Operations Centre

Hosting visualisation and analytics tools for COVID-19 studies

Hosting visualisation and analytics tools for COVID-19 studies

Exploring perceptions towards climate change over time on Twitter

Exploring perceptions towards climate change over time on Twitter

Coastal image classification and nalysis based on convolutional neural betworks and pattern recognition

Coastal image classification and nalysis based on convolutional neural betworks and pattern recognition

Calcium signalling in salivary gland acinar cells

Calcium signalling in salivary gland acinar cells

Anti-corruption regulations for promoting socially responsible practices

Anti-corruption regulations for promoting socially responsible practices

Determinants of translation efficiency in the evolutionarily-divergent protist Trichomonas vaginalis

Determinants of translation efficiency in the evolutionarily-divergent protist Trichomonas vaginalis

Analysing text data by time-series feature engineering

Analysing text data by time-series feature engineering

An investigation into Leap Motion device for “gesture-as-sign”

An investigation into Leap Motion device for “gesture-as-sign”

Antibiotic resistance and the “end of modern medicine ”

Antibiotic resistance and the “end of modern medicine ”

Develop short-term eruption warning systems for Whakaari and other volcanoes

Develop short-term eruption warning systems for Whakaari and other volcanoes

Evenly spaced observation fields from irregularly sampled data in the Southern Ocean

Evenly spaced observation fields from irregularly sampled data in the Southern Ocean

Measuring impact of entrepreneurship activities on students’ mindset, capabilities and entrepreneurial intentions

Measuring impact of entrepreneurship activities on students’ mindset, capabilities and entrepreneurial intentions

Using Zebra Finch data and deep learning classification to identify individual bird calls from audio recordings

Using Zebra Finch data and deep learning classification to identify individual bird calls from audio recordings

NETwork! analysis in cancer – managing genomics research data and building a repository workflow

NETwork! analysis in cancer – managing genomics research data and building a repository workflow

The Coronary Atlas – data processing workflow optimisation

The Coronary Atlas – data processing workflow optimisation

3D visualisation of indigenous burial site in Roonka

3D visualisation of indigenous burial site in Roonka

Automated measurement of intracranial cerebrospinal fluid volume and outcome after endovascular thrombectomy for ischemic stroke

Automated measurement of intracranial cerebrospinal fluid volume and outcome after endovascular thrombectomy for ischemic stroke

A new ‘stratigraphy’: interpreting object relationships with 3D point densities

A new ‘stratigraphy’: interpreting object relationships with 3D point densities

Towards the use of deep learning techniques for storm surge prediction

Towards the use of deep learning techniques for storm surge prediction

Using simple models to explore complex dynamics: A case study of macomona liliana (wedge-shell) and nutrient variations

Using simple models to explore complex dynamics: A case study of macomona liliana (wedge-shell) and nutrient variations

Development of Machine Learning methodology for genomic research

Development of Machine Learning methodology for genomic research

An Archaeological database for threatened North Island rock art in New Zealand

An Archaeological database for threatened North Island rock art in New Zealand

Presence: distributed mixed reality learning environment

Presence: distributed mixed reality learning environment

Digital video and the early learning lab

Digital video and the early learning lab

Publishing the Bay of Island Bottlenose dolphin catalogue

Publishing the Bay of Island Bottlenose dolphin catalogue

Modelling the diurnal cycle* of winds and clouds

Modelling the diurnal cycle* of winds and clouds

Presence: distributed mixed reality learning environment

Presence: distributed mixed reality learning environment

Using research virtual machines to analyse fMRI datasets

Using research virtual machines to analyse fMRI datasets

Genomic Virtual Lab (GVL) as a bioinformatics training platform

Genomic Virtual Lab (GVL) as a bioinformatics training platform

SwiftLaTeX- Exploring web-based true WYSIWYG editing for digital publishing

SwiftLaTeX- Exploring web-based true WYSIWYG editing for digital publishing

Climate change impacts on weather-related hazards

Climate change impacts on weather-related hazards

Understanding tumour evolution through augmented reality

Understanding tumour evolution through augmented reality

Myocardial motion tracking and strain calculation using Deep Learning networks

Myocardial motion tracking and strain calculation using Deep Learning networks

OnTask pilot at the Centre for Learning and Research in Higher Education

OnTask pilot at the Centre for Learning and Research in Higher Education

Visualising the University campus in 3D

Visualising the University campus in 3D

Visualising protein interaction

Visualising protein interaction

Biological heritage National Science Challenge eDNA virtual hub

Biological heritage National Science Challenge eDNA virtual hub

Interactive AR art – Project Gordon

Interactive AR art – Project Gordon

1-D numerical models of post-glacial river evolution

1-D numerical models of post-glacial river evolution

Mathematically modelling gastrointestinal electrical activity

Mathematically modelling gastrointestinal electrical activity

3D Cryo-EM reconstructions of macromolecular complexes

3D Cryo-EM reconstructions of macromolecular complexes

Engine knock in a spark-ignition engine with hydrogen supplementation

Engine knock in a spark-ignition engine with hydrogen supplementation

The complex unsteady flow within a fluid-filled annulus and its transition to turbulence

The complex unsteady flow within a fluid-filled annulus and its transition to turbulence

Using data mining for digital ink recognition

Using data mining for digital ink recognition

The landscape costs of brushtail possum dispersal

The landscape costs of brushtail possum dispersal

Accelerating the discovery of natural products made by orphan megasynthases

Accelerating the discovery of natural products made by orphan megasynthases

Improving the short term precipitation forecasts for New Zealand

Improving the short term precipitation forecasts for New Zealand

Finding genetic variants responsible  for human disease hiding in the universe of benign variants

Finding genetic variants responsible for human disease hiding in the universe of benign variants

Revealing key processes in enzyme efficiency through high performance computing

Revealing key processes in enzyme efficiency through high performance computing

3D Electromagnetic modeling and simulation using heterogeneous computing

3D Electromagnetic modeling and simulation using heterogeneous computing

Hemodynamics in the microcirculation

Hemodynamics in the microcirculation

Putting turbulence to work

Putting turbulence to work

Why are some molecules drugs?

Why are some molecules drugs?

Bayesian additive regression trees  vs logistic regression – estimation of propensity scores

Bayesian additive regression trees vs logistic regression – estimation of propensity scores

Fully coupled thermo-hydro-mechanical modelling of permeability enhancement by the finite element method

Fully coupled thermo-hydro-mechanical modelling of permeability enhancement by the finite element method

Modelling dispersal and ecological competition in a statistical phylogeographic framework

Modelling dispersal and ecological competition in a statistical phylogeographic framework

Studying the shape and the size of the universe

Studying the shape and the size of the universe

Planet hunting

Planet hunting

Simulating quantum mechanics on high performance computing cluster

Simulating quantum mechanics on high performance computing cluster

Multiscale modelling of saliva secretion

Multiscale modelling of saliva secretion

Modelling dual reflux pressure swing adsorption (DR-PSA) units for gas separation in natural gas processing

Modelling dual reflux pressure swing adsorption (DR-PSA) units for gas separation in natural gas processing

Improving the treatment of heart disease

Improving the treatment of heart disease

Estimating migration rates in the budding yeast Saccharomyces cerevisiae

Estimating migration rates in the budding yeast Saccharomyces cerevisiae

Number theoretic algorithms in cryptography

Number theoretic algorithms in cryptography

Molecular phylogenetics uses genetic data to reconstruct the evolutionary history of individuals, populations or species

Molecular phylogenetics uses genetic data to reconstruct the evolutionary history of individuals, populations or species

Phylogeny and phylogeography of the family kyphosidae (Perciformes: teleostei)

Phylogeny and phylogeography of the family kyphosidae (Perciformes: teleostei)

Testing what cosmic inflation really predicts

Testing what cosmic inflation really predicts

Multigene environmental DNA data analysis for New Zealand genomic observatory

Multigene environmental DNA data analysis for New Zealand genomic observatory

Finding genetic variants responsible for human disease hiding in universe of benign variants

Finding genetic variants responsible for human disease hiding in universe of benign variants

BEAST, Bayesian evolutionary analysis sampling trees

BEAST, Bayesian evolutionary analysis sampling trees

The formation of surface archaeological deposits in arid Australia

The formation of surface archaeological deposits in arid Australia

Statistical modelling of carryover effects after cessation of treatments

Statistical modelling of carryover effects after cessation of treatments

High-resolution cryo-electron microscopy of protein complexes and machines

High-resolution cryo-electron microscopy of protein complexes and machines

ARCI, archaeology eResearch collaboration initiative

ARCI, archaeology eResearch collaboration initiative

Optimisation of blades on large wind turbines with individual pitch control and trailing edge flaps

Optimisation of blades on large wind turbines with individual pitch control and trailing edge flaps

Quality of care and outcomes in children with cleft lip and/or palate

Quality of care and outcomes in children with cleft lip and/or palate

Geographic and temporal information retrieval on massive document collections

Geographic and temporal information retrieval on massive document collections

Homodynamics in the microcirculation

Homodynamics in the microcirculation

Processing structure-from-motion photogrammetry on the cluster

Processing structure-from-motion photogrammetry on the cluster

Computational investigation of catalysis mechanisms for polyurethane synthesis

Computational investigation of catalysis mechanisms for polyurethane synthesis

Virtual childhood obesity prevention laboratory

Virtual childhood obesity prevention laboratory

Giving Pacific research greater reach

Giving Pacific research greater reach

Development of novel waveguides  in the terahertz (THz) region

Development of novel waveguides in the terahertz (THz) region

Modelling of costs of diets  by INFORMAS

Modelling of costs of diets by INFORMAS

Foodback

Foodback

Finite element method code for  modelling biological cells

Finite element method code for modelling biological cells

The future of memory: Neuroimaging memory and imagination with functional MRI

The future of memory: Neuroimaging memory and imagination with functional MRI

Modelling and visualisation of calcium waves in parotid acinar cells

Modelling and visualisation of calcium waves in parotid acinar cells

Mapping donor contributions in the Pacific

Mapping donor contributions in the Pacific

Visualising humpback whale migration

Visualising humpback whale migration

Visualising the 2010 and 2011  Canterbury earthquakes

Visualising the 2010 and 2011 Canterbury earthquakes

Data management planning for MOA*

Data management planning for MOA*

Research data publishing  and preservation at COMPASS

Research data publishing and preservation at COMPASS

Centre for eResearch machine learning service

Centre for eResearch machine learning service

Building a discrete global  grid gazetteer service

Building a discrete global grid gazetteer service

The new Wanhal catalogue

The new Wanhal catalogue

Passive acoustic modelling

Passive acoustic modelling

Using GPUs to expand our understanding of the Solar System

Using GPUs to expand our understanding of the Solar System

Shedding new light on dark matter

Shedding new light on dark matter

Aerodynamics modelling paves the way for improved yacht designs

Aerodynamics modelling paves the way for improved yacht designs

Modernising models to help diagnose or treat disease and injury

Modernising models to help diagnose or treat disease and injury

Wandering around the molecular landscape: embracing virtual reality as a research showcasing outreach and teaching tool

Wandering around the molecular landscape: embracing virtual reality as a research showcasing outreach and teaching tool

ALTER: Between human and nonhuman – a VR art exhibition

ALTER: Between human and nonhuman – a VR art exhibition

Disposition of Microsoft HoloLenses for a Pop-Up Reality Shop to demonstrate the progress of a research project

Disposition of Microsoft HoloLenses for a Pop-Up Reality Shop to demonstrate the progress of a research project

Improving diagnosis for schistosomiasis by using the ‘metabolic footprint’ of urine samples from an animal model of Schistosoma infection to identify possible biomarkers

Improving diagnosis for schistosomiasis by using the ‘metabolic footprint’ of urine samples from an animal model of Schistosoma infection to identify possible biomarkers

Making stroke recovery prediction tools freely available

Making stroke recovery prediction tools freely available

MFT-ICR mass spectrometry data management and analysis workflow

MFT-ICR mass spectrometry data management and analysis workflow

Taking a ‘Big Data’ approach to find new clinical-omic associations in cancer

Taking a ‘Big Data’ approach to find new clinical-omic associations in cancer

Growing Up in New Zealand

Growing Up in New Zealand

Improving arrival time predictions for vehicles in a public transport network

Improving arrival time predictions for vehicles in a public transport network

Distributed and cloud-based control at field-level for systems interacting with soft bodies

Distributed and cloud-based control at field-level for systems interacting with soft bodies

Mobile Click Fraud Attack (MCFA)

Mobile Click Fraud Attack (MCFA)

Skin-omics: exploring the volatile organic compounds on human skin

Skin-omics: exploring the volatile organic compounds on human skin

New analytics tools for workload planning for the 2018 New Zealand Census

New analytics tools for workload planning for the 2018 New Zealand Census

Visualising the New Zealand Index of Multiple Deprivation

Visualising the New Zealand Index of Multiple Deprivation