Domain-informed and neural-optimized belief assignments: A framework applied to cultural heritage

Identifying pigments in Cultural Heritage artifacts is key to uncovering their origin and guiding conservation strategies. Although recent advances in non-invasive imaging have enabled the collection of rich multimodal data, existing methods often fall short in dealing with uncertain, ambiguous, or noisy information. This paper introduces a versatile fusion framework grounded in Belief Function Theory, combining domain-informed evidence modeling with neural optimization. Specifically, we propose a general strategy for assigning mass functions by leveraging expert knowledge encoded in parametric Evidence Mapping Functions, which are further refined through task-specific training using constrained neural networks. When applied to pigment classification, our method demonstrates robustness against source variability and class ambiguity. Experiments conducted on both synthetic and mock-up datasets validate its effectiveness and suggest promising potential for broader applications. • We propose a novel framework that combines domain-informed modeling and neural optimization for assigning belief masses. • Mass functions are derived using parametric Evidence Mapping Functions (EMFs) grounded in expert knowledge. • EMFs are implemented as constrained neural networks, enabling task-specific training under prior constraints. • The method is compatible with multimodal configurations, and was tested on pigment identification. • Extensive experiments on synthetic and real mock-up datasets show robustness to class ambiguity, noise, and data scarcity.