We are pleased to invite you to contribute to our sessions in the CAA2022 (Annual Conference of Computer Applications and Quantitative Methods in Archaeology) that will be held from Aug 3rd to 8th in Oxford (UK), in a hybrid format. GIAP-ICAC researchers are leading two sessions:
S13: Machine and deep learning methods in archaeological research: beyond site detection
Arnau Garcia-Molsosa, Hector A. Orengo, Iban Berganzo-Besga, Catalan Institute of Classical Archaeology
Although machine learning (ML) and deep learning (DL) related methods have been in use for several decades, they have only been applied to archaeological problems recently. Some early implementations focussed on the classification, seriation and analysis of material culture such as artistic representations (Barceló 1995a and 1995b, Di Ludovico and Ramazzotti 2005), use-wear of prehistoric tools (Van den Dries 1998), historical glass artifacts and ancient coins (Van der Maaten et al. 2007). The application of ML and DL in archaeology has experienced a strong turn towards the detection of archaeological sites during the last years (but see Wright and Gattiglia 2018 and Orengo and Garcia-Molsosa 2019 for the identification of ceramic fragments and Oonk and Spijker 2015 for geochemical analysis) making heavy use of multispectral satellite and lidar data. Since the pioneering work of Menze and Ur 2012, the wider availability of data (in particular high-resolution lidar), cloud computing platforms and AI processes and code has boosted ML and DL site-based detection (e.g. Lisset al. 2017, Trier et al. 2018, Berganzo-Besga et al. 2021).
Despite the importance of site location for the discipline, it is obvious that ML and DL enclose enormous potential to boost many areas of archaeological research, particularly within but not restricted to the field of computer vision. ML and DL are able to build inference models from sample data that can organise information without the need to explicitly program the process. Archaeologists are gaining skills and access to computational resources while, at the same time, new interfaces facilitate the use of these techniques to researchers not specialised in computer methods (see, e.g. Altaweel et al. 2022). The increased availability of models and examples opens the possibility to extend the debate towards how specific historical and archaeological debates can benefit from these new analytical instruments and the knowledge they generate.
This session aims to bring together archaeological ML and DL applications, discuss the problems related to their application and offer insight on to best practices. We welcome contributions about the application of ML and DL to different aspects of archaeological research and practice. With that perspective, we expect to provide a platform where the participants can observe and discuss the ensemble of opportunities that ML and DL can provide, with special interest on the possibility to create synergies between different fields of application, that are being developed in isolation.
Some of the suggested topics for the session are:
- Case studies on the application of AI to different sources of archaeological information. That can include the analysis of texts, artistic representations, bioarchaeological remains, material culture or archaeological sites. Combinations of such will be particularly welcome.
- Best practices and procedures, which can include comparative analysis. We are interest on examples on how to approach sensible datasets and how to facilitate reproducibility and Open Science principles in general.
- Big-data, data cleaning, data augmentation and data ingestion, as transversal challenges in many fields. Contributions addressing how researchers are developing, working with and taking advantage of large datasets, problems arising and potential solutions.
- The continuously increasing availability of detectors and methods makes sometimes difficult to select the best processes and algorithms for specific tasks. In this regard we welcome talks on algorithm selection, modification and performance evaluation.
- Talks addressing the development of computationally cost-effective workflows, in particular for the use or analysis of large datasets and the application of intensive computing processes will constitute a welcome addition to the session’s discussion.
S14: Large-scale and intensive computational workflows in archaeological remote sensing: from big data to data science
Francesc C. Conesa, Arnau Garcia-Molsosa, Hector A. Orengo, Catalan Institute of Classical Archaeology
The last few years have seen an unprecedented advance in the application of computational approaches for the remote analysis of archaeological landscapes, sites and features. This progress is mostly related to improvements in availability, diversity and quality of remote sensing data acquired from multiple platforms (ranging from satellite imagery to UAVs) and sensors (e.g. multispectral, radar, lidar, thermal) but also to increased access to high performance computing. This is partly related to the development of multi-petabyte catalogues of geospatial datasets linked to cloud computing environments accessible through web-based application programming interfaces associated to interactive development environments. These have granted the research community unparalleled access to geospatial data and computing power, and facilitated the development of large-scale, multi-temporal and multi-sensor analyses of the Earth’s surface. These environments have also been instrumental in the implementation of intensive computational processes, such as machine learning-based data classification, multi-scale topographic analysis, long-term time series analysis, and so on.
Many computing platforms have recently gain some popularity and are boosting a fast-growing number of applications. Among those, we highlight the of Amazon Web Services or the Data and Information Access Services that are run by the Copernicus satellite program. In particular, Google’s Earth Engine have been integrated into the archaeologist toolkit in a wide range of topics such as site detection and the long-term monitoring and management of cultural landscapes (see for instance Agapiou 2016; Rayne e al. 2017; Orengo and Petrie, 2018; Orengo 2020, Agapiou 2021).
This session will aim at showcasing and discussing new computational workflows for the treatment and analysis of large or complex remote sensing and other geospatial datasets. The adoption of cloud-computing platforms and other reproducible processing operations ultimately leave more time and resources for data interpretation and compared studies. We therefore encourage submissions that highlight innovative remote-based archaeological applications in the following or similar topics:
- Implementation of high-performance computing workflows
- Application of cloud-based computing platforms to archaeological remote sensing problems
- Use of multi-petabyte repositories of geospatial datasets
- Complex computing or multiplatform workflows
- Remote sensing analysis of large areas
- Landscape analyses over time-series
- Geospatial data cleaning and preparation
- Synergistic use of multispectral and radar satellite imagery
- Integration of multi-sensor or complex types of remote sensing data
For more details about the sessions, follow this link: https://2022.caaconference.org/sessions/
The call for abstracts is now open (deadline for submissions March 21st 2022). Link to the online submission process: https://2022.caaconference.org/call-for-papers-and-posters/
We look forward to seeing you in Oxford!
