12 August 2024 in IEEE Transactions on Intelligent Vehicles
LiDAR-based place recognition contributes significantly to identifying loop closure candidates for long-term simultaneous localization and mapping (SLAM) systems and obtaining global position for intelligent vehicles. Although the existing Transformer-based methods have achieved impressive results, they lack the capability of harnessing Transformers to establish long-range dependencies and suffer from high computational costs. To address these limitations, this study proposes a hierarchical range image transformer network named RangePlace to generate discriminative global descriptors for place recognition. By leveraging shifted window-based attention and depth-wise convolution, the proposed network can effectively and efficiently extract both global and local features. This not only enhances the network's capability to learn long-range information but also reduces the computational costs. In addition, a Pyramid Feature Mix module is incorporated to aggregate multi-scale feature maps to unified global descriptor, thus significantly enhancing place recognition performance. The proposed method is verified by experiments on the KITTI Odometry dataset, NCLT dataset, Ford Campus dataset and NuScenes dataset. The results demonstrate that the proposed method can achieve state-of-the-art performance in terms of recall rates. Moreover, it has good robustness to viewpoint yaw-rotation and strong generalization ability in various environments. Furthermore, it can be applied to real-time localization systems. The open-sourced codes used in this study are publicly available at https://github.com/JiLiBIT/RangePlace .
12 February 2026 in arXiv
Generative models have advanced significantly in realistic image synthesis, with diffusion models excelling in quality and stability. Recent multi-view diffusion models improve 3D-aware street view generation, but they struggle to produce place-aware and background-consistent urban scenes from text, BEV maps, and object bounding boxes. This limits their effectiveness in generating realistic samples for place recognition tasks. To address these challenges, we propose DiffPlace, a novel framework that introduces a place-ID controller to enable place-controllable multi-view image generation. The place-ID controller employs linear projection, perceiver transformer, and contrastive learning to map place-ID embeddings into a fixed CLIP space, allowing the model to synthesize images with consistent background buildings while flexibly modifying foreground objects and weather conditions. Extensive experiments, including quantitative comparisons and augmented training evaluations, demonstrate that DiffPlace outperforms existing methods in both generation quality and training support for visual place recognition. Our results highlight the potential of generative models in enhancing scene-level and place-aware synthesis, providing a valuable approach for improving place recognition in autonomous driving