TY - JOUR
T1 - Reconstructing rock art chronology with transfer learning
T2 - A case study from Arnhem Land, Australia
AU - Kowlessar, Jarrad
AU - Keal, James
AU - Wesley, Daryl
AU - Moffat, Ian
AU - Lawrence, Dudley
AU - Weson, Abraham
AU - Nayinggul, Alfred
AU - Mimal Land Management Aboriginal Corporation,
PY - 2021
Y1 - 2021
N2 - In recent years, machine learning approaches have been used to classify and extract style from media and have been used to reinforce known chronologies from classical art history. In this work we employ the first ever machine learning analysis of Australian rock art using a data efficient transfer learning approach to identify features suitable for distinguishing styles of rock art. These features are evaluated in a one-shot learning setting. Results demonstrate that known Arnhem Land Rock art styles can be resolved without knowledge of prior groupings. We then analyse the activation space of learned features and report on the relationships between styles and arrange these classes into a stylistic chronology based on distance within the activation space. By generating a stylistic chronology, it is shown that the model is sensitive to both temporal and spatial patterns in the distribution of rock art in the Arnhem Land Plateau region. More broadly, this approach is ideally suited to evaluating style within any material culture assemblage and overcomes the common constraint of small training data sets in archaeological machine learning studies.
AB - In recent years, machine learning approaches have been used to classify and extract style from media and have been used to reinforce known chronologies from classical art history. In this work we employ the first ever machine learning analysis of Australian rock art using a data efficient transfer learning approach to identify features suitable for distinguishing styles of rock art. These features are evaluated in a one-shot learning setting. Results demonstrate that known Arnhem Land Rock art styles can be resolved without knowledge of prior groupings. We then analyse the activation space of learned features and report on the relationships between styles and arrange these classes into a stylistic chronology based on distance within the activation space. By generating a stylistic chronology, it is shown that the model is sensitive to both temporal and spatial patterns in the distribution of rock art in the Arnhem Land Plateau region. More broadly, this approach is ideally suited to evaluating style within any material culture assemblage and overcomes the common constraint of small training data sets in archaeological machine learning studies.
KW - Arnhem Land
KW - machine learning
KW - Rock art
KW - style
UR - http://www.scopus.com/inward/record.url?scp=85103395622&partnerID=8YFLogxK
UR - http://purl.org/au-research/grants/ARC/DE170101447
UR - http://purl.org/au-research/grants/ARC/DE160100703
U2 - 10.1080/03122417.2021.1895481
DO - 10.1080/03122417.2021.1895481
M3 - Article
AN - SCOPUS:85103395622
SN - 0312-2417
VL - 87
SP - 115
EP - 126
JO - Australian Archaeology
JF - Australian Archaeology
IS - 2
ER -