Scanning Samian pottery during lockdown

Guest blog written by the Arch-I-Scan project, University of Leicester.

Although the second national lockdown meant that the Vindolanda fort and museum was required to close to the public from November 5^th to December 2^nd, the continuation of research activities was permitted. And so, at the beginning of the month, armed with our PPE and copious amounts of hand sanitiser, the Arch-I-Scan project team made our way from Leicester to Vindolanda for three weeks of pottery scanning on-site.

 

 Arch-I-Scan team members take a lunch break at Doncaster station, en route to Vindolanda.

‘Arch-I-Scan: Automated recording and machine learning for collating Roman ceramic tablewares and investigating eating and drinking practices’ is an AHRC-funded research project, led by Professor Penelope Allison from the School of Archaeology and Ancient History, University of Leicester, and Professor Ivan Tyukin from the University’s School of Mathematics and Actuarial Science. This project, which began in November 2019, is in the process of developing a state-of-the-art image-recognition and machine-learning service which can record complete/near complete vessels and more fragmentary remains, as well as digitally collate and store large quantities of data.

Arch-I-Scan will first compile a large dataset of images which detail aspects of size, shape, design and texture of pottery sherds of mainly Samian ware. The aim is to take hundreds of thousands of such images using handheld devices such as mobile phones and which are fed into the service as part of its ‘training’. Machine-learning operates on the premise that the more the service ‘sees’, the more it learns; in a sort of science-fiction-meets-archaeology sense, it is a process that utilises artificial intelligence algorithms similar to those of facial recognition software. It is projected that when sufficient data has been fed into the Arch-I-Scan software, it will begin to automatically recognise and record details of pottery remains. A proof-of-concept experiment was successfully carried out for the research network, 'Big Data on the Roman Table', also funded by the AHRC. The service which the Arch-I-Scan project is developing has the potential to revolutionise how these artefacts are identified and recorded, especially since it could be operated by non-specialists and specialists alike using handheld devices. The time saved in the initial identification process would result in more time left for expert analysis, which would ultimately facilitate greater, more nuanced understandings of the past. See more on the project aims here.

As described above, in order to ‘train’ the machine-leaning service, the Arch-I-Scan team will need to gather an immense body of data in the form of sherd images, which is precisely where Vindolanda comes in! We are collaborating with a number of project partner institutions – the Vindolanda Trust, Museum of London, Museum of London Archaeology, University of Leicester Archaeology Service (ULAS) and Colchester and Ipswich Museum Service – which have provided access to their extensive collections of Roman fineware remains. In addition to access to their Roman pottery collections and curatorial support, the Vindolanda Trust provided our project team members with accommodation on-site for three weeks.

 

 Some examples of photos taken by the Arch-I-Scan team of the Samian ware sherds at Vindolanda.

It has been truly surreal to have the Roman fort of Vindolanda to ourselves (or at least share with Vindolanda staff) over the past three weeks during the national lockdown. From our accommodation space near the museum, we walked through the fort (best commute ever!) to the brand-new Robin Birley Archaeology Centre where we completed our daily scanning activities. We were the first research team to utilise the new built-for-purpose Centre since it arrived at Vindolanda in March 2020, and we all enjoyed working in the large, open space, which looked out onto the archaeological site.  When we originally planned our scanning session at Vindolanda, we had been looking forward to sharing the centre with volunteers, but unfortunately the current pandemic meant that we would have to limit ourselves to just be a few core Arch-I-Scan team members, with each of us spaced out more than 3 metres apart. Still, even social distancing could not keep us from lively banter while we worked, and since this was our first scanning trip since the initial lockdown in March, our spirits were high. It was just so good to be together again in-person and back in sherd-scanning action - and we were beyond thrilled to handle examples of Samian ware pottery from the 2013-2015 Vindolanda excavation seasons.

The Robin Birley Centre was our centre of scanning operations for three weeks. It was so nice to work with the pottery in a building which overlooked the archaeological site!

The process of ‘scanning’ – photographing – our sherds consists of our taking six photos which cover each face of the sherd, as well as (when possible) a photo of a magnified fresh break. The AI has the potential to see these images, identify thousands of features, and between these, in theory find patterns in pottery better than the human eye. Currently, we are in the stage of collecting as many images as possible from which it will be designed to identify patterns. At Vindolanda, informal competitions for who could collect the most images in a day kept things lively (Santos Nuñez, our Maths post-doc, holds the current team record!)

Although the actual scanning process at Vindolanda was repetitive, it was enjoyable, and being on-site right next to a Roman fort while we worked made the experience particularly exciting. Because we were unable to travel back to Leicester at the weekends, we took opportunities to go on long walks along the Hadrian’s Wall path (pro-tip: the stretch of the Wall near to Vindolanda is arguably the best bit).

 

 We enjoyed long walks along the Hadrian’s Wall national trail at the weekends.

We are immensely grateful to the staff members at the Vindolanda Trust who made this research visit possible, and who lent us whatever help that we required whilst on-site. We look forward to hopefully returning early in the new year to continue our data collection of the Vindolanda Samian ware materials. Principal Investigator, Penelope Allison, was unable to join the team for scanning this time but writes, ‘It is great that we have been able to move this project forward in this way at a time when such research activities, usually involving members of the wider community, are currently being severely curtailed. I am very grateful to all the team for their commitment to the project here and for great support from Andy Birley and the Vindolanda Trust team.’

For week-by-week details on our time at Vindolanda, as well as information on our other work over the past year, you can also check out the Arch-I-Scan project’s blog here.