ACM Journal on Computing and Cultural Heritage最新文献

Iconclass is an iconographic thesaurus, which is widely used in the digital heritage domain to describe subjects depicted in artworks. Each subject is assigned a unique descriptive code, which has a corresponding textual definition. The assignment of Iconclass codes is a challenging task for computational systems, due to the large number of available labels in comparison to the limited amount of training data available. Transfer learning has become a common strategy to overcome such a data shortage. In deep learning, transfer learning consists in fine-tuning the weights of a deep neural network for a downstream task. In this work, we present a deep retrieval framework, which can be fully fine-tuned for the task under consideration. Our work is based on a recent approach to this task, which already yielded state-of-the-art performance, although it could not be fully fine-tuned yet. This approach exploits the multi-linguality and multi-modality that is inherent to digital heritage data. Our framework jointly processes multiple input modalities, namely, textual and visual features. We extract the textual features from the artwork titles in multiple languages, whereas the visual features are derived from photographic reproductions of the artworks. The definitions of the Iconclass codes, containing useful textual information, are used as target labels instead of the codes themselves. As our main contribution, we demonstrate that our approach outperforms the state-of-the-art by a large margin. In addition, our approach is superior to the M³P feature extractor and outperforms the multi-lingual CLIP in most experiments due to the better quality of the visual features. Our out-of-domain and zero-shot experiments show poor results and demonstrate that the Iconclass retrieval remains a challenging task. We make our source code and models publicly available to support heritage institutions in the further enrichment of their digital collections.

The existing methods of Facial Expression Recognition (FER) primarily analyze six basic expressions, namely, surprise, happiness, anger, sadness, fear, and disgust. The Indian performing arts use three more well-defined expressions—peaceful, proud, and erotic. This study proposes an intelligent dual-level expression evaluation system that classifies performance-specific expressions into nine classes, assigns intensity level to the expression, and suggests modifications to the user for precise exhibition of an expression. At decision level-1 of a dual-level system, an 11-state model is designed to classify the nine expressions. The model is verified using the Colored Petri Net that helps analyze the rules used for the classification. Decision level-1 is also implemented using input feature database and SVM classifier, which yields 95.77% accuracy. Further, at decision level-2, SVM is used to assign an intensity level to the correctly classified images. In case of incorrectly exhibited expressions, feedback is provided to the user about the incorrect facial component state. The application-specific image dataset is used for the present study. The qualitative comparison with the other FER approaches is also carried out. With the increasing popularity of Indian classical dance in Western and Asian countries, the dual-level system enables learners of performing arts to practice, evaluate, and improvise their expression skills.

Artworks need to be constantly monitored to check their state of conservation and to quickly spot the eventual presence of alterations or damages. Preventive conservation is the set of practices employed to reach this goal. Unfortunately, this results generally in a cumbersome process involving multiple analytical techniques. Consequently, methods able to provide a quick preliminary examination of the artworks (e.g., optical monitoring) seem very promising to streamline preventive conservation procedures. We are especially interested in the study of historical wood musical instruments, a kind of artwork particularly subject to mechanical wear since they are both held in museums and also occasionally played in concerts. Our primary goal is to detect possible altered regions on the surface of the instruments early and thus provide the experts some precise indications on where to apply more in-depth examinations to check for potential damages. In this work, we propose an optical monitoring method based on the a-contrario probabilistic framework. Tests were conducted on the “Violins UVIFL imagery” dataset, a collection of UV-induced fluorescence image sequences of artificially altered wood samples and violins. Obtained results showed the robustness of the proposed method and its capability to properly detect altered regions while rejecting noise.

Rural areas in Europe represent outstanding examples of Cultural and Natural Heritage (CNH) that could be used as a valuable asset for social and economic development. This article describes the process for developing a monitoring platform based on Key Performance Indicators (KPI) and implemented in six rural areas around Europe. The goal of this monitoring system is to provide evidence of the role of CNH in rural areas as a driver for sustainable growth. Several data collection procedures are described, including regular, non-regular, and co-monitoring. To combine the selected cross-thematic and multi-scale KPIs, weights have been assigned to indicators, according to the knowledge provided by domain experts and using group decision-making techniques. A detailed description of the dashboards developed for the monitoring platform, and all the information gathered is included. Several dashboards have been designed focusing on KPI values and their evolution.