Li Jiaqi | Searcharxiv

arXiv2024

Yuan: Research on the Concept of Digital World Analogue Scientific Infrastructure and Science Popularization Communication Based on Suzhou Gardens Pattern

In the current digital era, high security relies significantly on advanced concepts such as native security. However, the design and implementation of these concepts face challenges in enterprises and organizations. Leveraging advancements in Large Language Models (LLMs), we draw inspiration from the design principles of Suzhou Gardens, a UNESCO World Heritage site. By examining its core features, which align closely with those of the AI world simulator Sora, we extract three concurrent concepts to enhance the security of future digital infrastructures.We propose three guiding principles to facilitate the preliminary construction of the "Space Spider," a hyper-large scientific infrastructure. These principles will steer the development of the "Yuan" digital garden, establishing a "Chinese Series" focused on the construction pathways of the Yuan AI world simulator. The initial pilot of Yuan is expected to generalize various hyper-large scientific infrastructure scenarios, ultimately expanding into numerous high-security digital applications.Through the design concept of Suzhou Gardens, we aim to promote science communication and talent training in the field of cybersecurity. With the support of Yuan, we intend to extend our efforts to various digital construction domains. This initiative is poised to contribute significantly to the future of digital world simulators, emphasizing the integration of hyper-large scientific infrastructure with science communication and research dissemination.

cs.CR↗

arXiv2021

MVM3Det: A Novel Method for Multi-view Monocular 3D Detection

Monocular 3D object detection encounters occlusion problems in many application scenarios, such as traffic monitoring, pedestrian monitoring, etc., which leads to serious false negative. Multi-view object detection effectively solves this problem by combining data from different perspectives. However, due to label confusion and feature confusion, the orientation estimation of multi-view 3D object detection is intractable, which is important for object tracking and intention prediction. In this paper, we propose a novel multi-view 3D object detection method named MVM3Det which simultaneously estimates the 3D position and orientation of the object according to the multi-view monocular information. The method consists of two parts: 1) Position proposal network, which integrates the features from different perspectives into consistent global features through feature orthogonal transformation to estimate the position. 2) Multi-branch orientation estimation network, which introduces feature perspective pooling to overcome the two confusion problems during the orientation estimation. In addition, we present a first dataset for multi-view 3D object detection named MVM3D. Comparing with State-Of-The-Art (SOTA) methods on our dataset and public dataset WildTrack, our method achieves very competitive results.

cs.CV↗