EyCon (Visual AI and Early Conflict Photography)

As more and more museums and archives are making their image collections digitally available, new forms of knowledge and exploration of this growing mass of pictures from the past are needed. Computational techniques, which can process and help visualise tens of thousands of images, will be key to unlocking these digital archives in the future. One of the main achievements of the project is to create an innovative database on wartime photography that will include previously unpublished material. EyCon focuses on early conflict photography (1890-1918) documenting mass armed violence, from often overlooked colonial campaigns to the First World War. It is designed as a collection of thousands of original photographs and printed images: despite its scale, this collection will be made discoverable thanks to computational techniques. One of EyCon's tasks is to improve the recognition of objects in images that are very different from the photographs current models have trained on. EyCon will also provide and apply tools to identify similarities between photographs and printed images in contemporary prints in order to retrace the circulation of images. Various visualisations of the entire corpus will also be developed and tested to help users navigate a very large collection of images in a rewarding way.

The large-scale discovery of early wartime imagery raises a wide range of ethical and methodological issues that EyCon will directly address.

First, EyCon will develop solutions to overcome bias and inaccuracies when AI techniques are applied to a potentially contested visual heritage. Computational methods are not neutral. They may reproduce existing power structures or fail to unravel the unbalanced relationships of the past. EyCon will take great care in emphasising both the benefits and limitations of distant vision.

Second, wartime photographs can be very sensitive material, especifically in colonial and imperial contexts that are characterised by racial and gendered violence. Making these pictures available and discoverable in a sensible way that might open the way for a multiplicity of narratives and perspectives is central to the project.

Third, while EyCon will help restore the visibility of several conflicts in thousands of images, it also has to consider absent images, forms of violences and conflicts that were not recorded by cameras and that can be crushed under the weight of a large mass of digitised visual material.

The EyCon project will lead to the following research outputs:

_2 workshops organised in collaboration with Project Partners
_2 collections of essays (special issues of journals or edited volumes)
_1 demo website integrating all AI functionalities and a metasearch engine to explore data held on Huma-Num and partner institutions' servers. This website will be designed for both specialist and non-specialist end-users. It will be conceived as a user-friendly tool for everyone with or without technical expertise. It will allow us to offer increased discoverability and smart visualisations to users who want to explore the entire Eycon corpus.
_An open-source repository (Gitlab) that will make EyCon's scripts and other tools openly available to a broad community. This repository will be useful for data scientists and developers to reuse the scripts and data produced by the project. This will ensure that the project code is replicable and usable in other projects.

Archives are of course not reserved to academic researchers. The demo website will foster public engagement on the topic of early conflict photography and computational methods applied to the history of photography. Associated social media and a dedicated Eycon list-serv will help us connect with interested parties - in academia, archival institutions and beyond.