IPALIA: Improving Public Accessibility of Large Image Archives

IPALIA is a multidisciplinary project which aims at improving public accessibility of large image databases and collections. We do this by development of an AI based automatic annotation method that can generate high-quality descriptions of the contents of photographs.

The EU Accessibility Directive came into effect in 9/2020 that requires textual descriptions of image contents to be added to all public images on the Web. Currently there are no inexpensive and efficient methods for these purposes and this requirement appears as the biggest bottleneck hindering the public service providers to fulfill the directive.

In TAU/ITC three areas contribute to this project: machine learning, information retrieval and visual studies. We collaborate with the Finnish Defence Forces Photograph Archive and Tampere Museum Centre Vapriikki’s Image archive. Primary material for the project is the Finnish Wartime Photograph Archive (FWPA). FWPA consists of almost 160,000 photographs. Most of the pictures were taken by Information Company (IC) photographers in 1939-1945.

Our approach for generating descriptions has some unique features - such as taking into account the spatial qualities of the photographs. In 2021 we have developed a new tool, Automatic Image Content Extraction (AICE). AICE is a groundbraking framework as it provides a general model for analyzing and searching large image collections. It can be applied in various domains in humanities and social sciences, and it can be adjusted and scaled in to various research settings.

Visual studies

Anssi Männistö, anssi.mannisto [at] tuni.fi

Machine learning team

Jenni Raitoharju, jenni.raitoharju [at] tuni.fi

Mert Seker, mert.seker [at] tuni.fi

Information retrieval team

Sanna Kumpulainen, sanna.kumpulainen [at] tuni.fi

Hille Ruotsalainen, hille.ruotsalainen [at] tuni.fi