As a subject area of symmetry, multiple instance learning (MIL) is a special form of a weakly supervised learning problem where the label is related to the bag, not the instances contained in it. The difficulty of MIL lies in the incomplete label information of instances. To resolve this problem, in this paper, we propose a novel diverse density (DD) and multiple part similarity combination method for multiple instance learning, named MILDMS. First, we model the target concepts optimization with a DD function constraint on positive and negative instance space, which can greatly improve the robustness to label noise problem. Next, we combine the positive and negative instances in the bag (generated by hand-crafted and convolutional neural network features) with multiple part similarities to construct an MIL kernel. We evaluate the proposed approach on the MUSK dataset, whose results MUSK1 (91.9%) and MUSK2 (92.2%) show our method is comparable to other MIL algorithms. To further demonstrate generality, we also present experimental results on the PASCAL VOC 2007 and 2012 (46.5% and 42.2%) and COREL (78.6%) that significantly outperforms the state-of-the-art algorithms including deep MIL and other non-deep MIL algorithms.
WaDang recognition can provide valuable information for digital protection of cultural relics. However, WaDang recognition is challenging due to a great number of variations, such as different characters, different background pattern, and different scale and rotation. In this paper, we try to solve these problems by using hand-crafted and deep features and integrate them into multiple instance learning, in which the features of instances are divided into two parts and their similarity is computed by multiple kernel learning. The experiments on our collected WaDang images, MUSK, and COREL datasets show that the proposed algorithm is effective and its performance is compared with other state-of-art algorithms.
The Acrobat Figure (Baixi) unearthed in the Qin Mausoleum's K9901 burial pit, vividly reflects the acrobat arts and other court entertainment activities of the Qin Dynasty and shows the image of acrobat artists at that time. The acrobatic figures provide a brand-new perspective for future generations to understand the society of the Qin Dynasty, with important historical, artistic, and scientific value. Currently, the digitalization researches of the Terracotta Army are mainly focused on Warriors’ burial pits and cultural relics, there is little attention on the digitalized display and dissemination of acrobatic figures. Since the wide applications of virtual reality (VR) technology in cultural heritage digitalization, this paper aims to conduct research on cultural experience and knowledge dissemination of the Acrobat Figure via VR. Firstly, a learner-centric progressive design method was adopted to virtually reconstruct acrobatic figures of the Qin Mausoleum with its environment, and a VR Baixi experience system was built; secondly, to evaluate the effectiveness of the cultural knowledge dissemination and learning, a comparative study was conducted between the VR Baixi and the traditional Baixi documentary; finally, the users’ experience data on knowledge dissemination results (KDR), performance expectancy (PE), and behavioral intention (BI) were collected and analyzed. The experimental results show that compared with the traditional approach, VR Baixi significantly improves the users' motivation and willingness to explore, and enhances their knowledge experience of the cultural heritage of the Acrobat Figure, which provides an effective solution for the knowledge dissemination and learning of cultural heritage on general audiences, especially among younger groups.
Historical garden sites are often characterized by substantial losses of above-ground remains,fragmented historical evidence,and uncertainty in reconstructing their former spatial and landscape configurations.Existing approaches to digital heritage reconstruction tend to emphasize the integration of geometric data,while systematic procedures for evaluating heterogeneous evidence and translating its reliability into reconstruction decisions remain limited.Visitor perception is likewise more commonly used to assess digital outcomes than to inform interpretative priorities.To address these limitations,this study develops an evidence-informed framework for the digital reconstruction of historical gardens that integrates evidence collection,reliability assessment,cross-validation,reconstruction constraints,and interpretative representation within a coherent workflow.Evidence sources are assigned different reconstruction roles according to their reliability and epistemic status.Archaeological,historical, and spatial evidence provide constraints for reconstructing landscape form and spatial organization,whereas visitor perception is treated as supplementary interpretative evidence for identifying publicly perceived landscape values and informing priorities for digital presentation.The framework was applied to the Tang Jiucheng Palace garden site through the integration of archaeological records,historical documents,spatial survey data,and visitor perception data.Key landscape elements,including architecture,water features,vegetation,and spatial organization,were identified and reconstructed,and an immersive digital environment combining panoramic navigation and virtual reality (VR) visualization was developed.The results indicate that the framework enhances the transparency and traceability of evidence use in digital reconstruction and supports more explicit interpretation of uncertainty and landscape values.The proposed approach provides a structured basis for the digital conservation,interpretation,and public communication of historical garden sites where above-ground remains are incomplete and evidence is heterogeneous.
Taihu stone, a natural limestone from the Taihu Lake basin in China, is characterized by perforations and irregular geometries formed through long-term water erosion and dissolution. Widely used in classical Chinese gardens, its morphological evolution remains quantitatively underexplored. This study examines Taihu stones from the Sui–Tang, Song–Yuan, and Ming–Qing periods based on point-cloud data obtained through multi-source 3D acquisition. Key parameters, including height, perforation characteristics, volumetric proportions, and morphological complexity, were quantitatively analyzed. The results show a morphological shift from relatively massive to more slender forms, increased perforation development, and greater morphological complexity. A digital morphological atlas of Taihu stones was then constructed to support cross-period comparison and digital reconstruction. This study provides a quantitative basis for typological analysis, identification, digital reconstruction, and heritage conservation of Taihu stones, and offers a reference for digital research on garden stones and cultural heritage.