Feature embedding has become a cornerstone technology for processing high-dimensional and complex data, which results in that Embedding as a Service (EaaS) models have been widely deployed in the cloud. To protect the intellectual property of EaaS models, existing methods apply digital watermarking to inject specific backdoor triggers into EaaS models by modifying training samples or network parameters. However, these methods inevitably produce detectable patterns through semantic analysis and exhibit susceptibility to geometric transformations including rotation, scaling, and translation (RST). To address this problem, we propose a novel fingerprinting framework called POSTER for EaaS models, rather than merely refining existing watermarking techniques. Different from watermarking, the proposed POSTER establishes EaaS model ownership through geometric analysis of embedding space's topological structure, rather than relying on the modified training samples or triggers. The key innovation lies in modeling the victim and suspicious embeddings as point clouds, allowing us to perform robust spatial alignment and similarity measurement, which inherently resists RST attacks. Experiments evaluated on visual and textual embedding tasks verify the superiority and applicability. This work reveals inherent characteristics of EaaS models and provides a promising solution for ownership verification of EaaS models under black-box scenarios.

Simultaneous Localization and Mapping (SLAM) has become a critical technology for intelligent transportation systems and autonomous robots and is widely used in autonomous driving. However, traditional manual feature-based methods in challenging lighting environments make it difficult to ensure robustness and accuracy. Some deep learning-based methods show potential but still have significant drawbacks. To address this problem, we propose a feature-based visual SLAM system based on the LightGlue deep learning network. It uses deep local feature descriptors to replace traditional hand-crafted features and a more efficient and accurate deep network to achieve fast and precise feature matching. Thus, we use the robustness of deep learning to improve the whole system. We have combined traditional geometry-based approaches to introduce a complete visual SLAM system for monocular, binocular, and RGB-D sensors. We thoroughly tested the proposed system on four public datasets: KITTI, EuRoC, TUM, and 4Season, as well as on actual campus scenes. The experimental results show that the proposed method exhibits better accuracy and robustness in adapting to low-light and strongly light-varying environments than traditional manual features and deep learning-based methods. It can also run on GPU in real time.

Visual Simultaneous Localization and Mapping (VSLAM) has undergone gradual development and found widespread application. However, existing VSLAM systems predominantly rely on static environment assumptions, leading to diminished robustness and localization accuracy in the presence of dynamic elements. Previous research has primarily employed geometric and semantic constraints to address dynamic regions of the scene. Nevertheless, their efficacy is limited in complex dynamic scenarios involving non-rigid objects, non-predefined motion targets, and low dynamic motion targets. Furthermore, the majority of dynamic SLAM methods are predominantly designed for indoor RGBD environments, resulting in a lack of generalizability. In this paper, a dynamic SLAM method that combines instance segmentation and optical flow called RSO-SLAM is proposed. RSO-SLAM is designed to operate effectively in diverse complex motion scenarios, both indoors and outdoors, and supports various visual sensor modes, including monocular, stereo, and RGBD setups. The proposed approach amalgamates semantic information and optical flow data by employing a “KMC:k-means$+$connectivity” based algorithm for motion region detection within the scene. Furthermore, it integrates an optical flow attenuation propagation strategy to facilitate meticulous motion probability computations and inter-frame propagation within each identified region. Our methodology’s superiority over existing dynamic SLAM approaches is firmly established through comprehensive evaluations across a diverse range of intricate dynamic scenarios. These evaluations encompass various conditions of high and low dynamism in both indoor and outdoor environments, accompanied by rigorous ablation experiments and real-world assessments. RSO-SLAM exhibits enhanced robustness and higher localization accuracy, rendering it well-suited for nearly all dynamic environments.