Researchers have developed a novel probabilistic method for dating historical manuscripts using visual script features. This approach treats manuscript dating as an evidential deep regression problem on a continuous year axis, enabling a neural network to output a predictive distribution with decomposed uncertainty. The model, combining an EfficientNet-B2 backbone with a Normal-Inverse-Gamma head, achieved a mean absolute error of 5.4 years on the DIVA-HisDB benchmark, significantly outperforming previous methods and offering better calibration with lower inference costs. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Introduces a novel AI approach for historical document analysis, potentially improving the accuracy and efficiency of dating manuscripts.
RANK_REASON Academic paper detailing a new methodology for manuscript dating using AI.