Datasets

Heritage datasets 

• NeRFBK: a collection of image sets primarily for NeRF / GS testing and evaluation

• Arch - A Benchmark for large-scale heritage point cloud segmentation: 17 annotated 3D point cloud scenes to test and evaluate 3D classification methods

•  Historical maps: some 19th century maps, in raster format, in need of vectorization (i.e. find buildings, streets, etc.)

Call for datasets

We invite submissions and presentation of datasets that can support the development and evaluation of AI-driven approaches for the classification, recognition and analysis of cultural heritage scenarios and objects. Cultural heritage presents unique challenges for computational methods, requiring models that can capture the complexity of material degradation, architectural forms, iconographic elements, and visual or morphological similarities across artefacts and structures. The availability of high-quality, well-documented datasets is essential to advancing AI research in this field.

We encourage dataset submissions that address, but are not limited to, the following themes:

• Material Degradation Recognition: Datasets capturing patterns of deterioration in heritage materials (e.g., erosion, biological growth, structural damage) to support automated analysis and predictive modeling.

• Architectural Form Recognition: Data collections focusing on the classification of architectural styles, structural components, or decorative elements in historical buildings and archaeological sites.

• Visual and Morphological Similarity Analysis: Datasets enabling the study of shape and texture similarities between artefacts, fragments, or structures to support digital reassembly, restoration, and stylistic analysis.

• Historical Object Classification: Datasets supporting the identification and classification of objects, inscriptions, or iconography within a heritage context.

• Multi-Modal Heritage Data: Collections integrating different sensing technologies (e.g., LiDAR, photogrammetry, multispectral imaging) to enhance AI-driven heritage documentation and analysis.

Submission Guidelines

Papers describing the proposed datasets should at least include the following:

1. Dataset description: A clear overview of the dataset, including its scope, sources, and intended use cases

2. Technical specifications: Details on data format, size, resolution, and any pre-processing applied

3. Scientific relevance: Explanation of how the dataset contributes to key challenges in cultural heritage research and AI applications

4. Ethical considerations: Information on data provenance, intellectual property rights, and any access restrictions.

Selected datasets will be featured in the workshop to foster discussion on best practices, challenges, and opportunities for AI in cultural heritage research.