The paper presents a lightweight 3D U‑Net designed to detect ultra‑sparse LiDAR occupancy of bat flight paths in nocturnal field recordings. By preserving temporal resolution and combining weighted binary cross‑entropy with Dice loss, the model overcomes the class imbalance that hampers standard reconstruction methods. Experiments on real LiDAR data, cross‑checked with acoustic monitoring, show that the U‑Net successfully recovers coherent occupancy patterns along bat trajectories, offering a practical foundation for large‑scale validation, clustering of flight tracks, and integration into biodiversity‑aware turbine curtailment strategies.
arXiv:2606. 09634v1 Announce Type: cross Abstract: 3D object detection is the backbone of perception for automated vehicles (AV) and broader intelligent transportation systems applications.
3D object detection is the backbone of perception for automated vehicles (AV) and broader intelligent transportation systems applications. Long-range detection is challenging because sensing evidence is sparse; yet this ``long-range'' scenario is routine in traffic.
DXPR is a depth‑based cross‑modal place recognition framework that matches monocular camera queries to a LiDAR map using a single vision foundation model backbone. By converting both modalities into a unified depth image representation, DXPR learns modality‑invariant global descriptors without modality‑specific encoders. A geometry‑aware overlap miner refines pairwise metric learning by computing pixel‑level overlap scores, and extensive tests on KITTI and Boreas show strong performance across seasons, weather, and day/night conditions, outperforming prior CMPR baselines.
arXiv:2606. 32023v1 Announce Type: cross Abstract: Forest attributes are essential for national-scale resource monitoring.
SARFusion introduces a scene-aware routing approach for camera‑LiDAR 3D object detection, decoupling object‑query decoding into separate camera, LiDAR, and fusion branches. By estimating a global scene reliability prior and incorporating object‑level evidence, each query is routed to the most suitable branch, reducing cross‑modal interference. The method achieves strong performance on the nuScenes test set (72.5 mAP, 74.4 NDS) and demonstrates robustness to sensor corruptions and environmental changes.
arXiv:2606. 20189v3 Announce Type: replace-cross Abstract: Leveraging Vision Foundation Models (VFMs) for camera-to-LiDAR knowledge distillation offers a promising solution to the scarcity of annotated data needed to represent the immense geometric and kinematic diversity of real-world autonomous driving (AD).
arXiv:2606. 20189v1 Announce Type: cross Abstract: Leveraging Vision Foundation Models (VFMs) for camera-to-LiDAR knowledge distillation offers a promising solution to the scarcity of annotated data needed to represent the immense geometric and kinematic diversity of real-world autonomous driving (AD).
KSG‑Net introduces a Key‑Sparse and Global‑Context learning framework for maritime 3D ship detection, addressing weak feature representation of small, sparse vessels and limited global modeling of large vessels. The network employs a Key Sparse Multi‑scale Aggregation module to select informative voxels and aggregate cross‑scale features, and a Global Context Aggregation module to capture long‑range geometric dependencies via scene‑level context modeling. Experiments on the Thames River vessel dataset and simulated data show that KSG‑Net outperforms existing methods in multi‑scale vessel detection and remains robust in complex maritime environments.
arXiv:2609.38864v1 Announce Type: new Abstract: Embodied tasks demand accurate, flexible, and semantically rich 3D scene representations. 3D semantic occupancy is well suited to this requirement, as...
arXiv:2607. 04541v1 Announce Type: cross Abstract: Camera-radar (CR) fusion is a practical sensing configuration for autonomous driving, but existing models are typically trained with task-specific supervision, limiting reusable representation learning.
GEM is a Generative LiDAR world model that uses a deformable Mamba architecture to better handle the disorder of LiDAR point clouds and distinguish dynamic objects from static structures. The model tokenizes LiDAR sweeps, unsupervisedly disentangles dynamic and static features, and applies a tri‑path deformable Mamba for selective scanning and adaptive gating fusion, improving spatial‑temporal understanding. Experiments show GEM outperforms existing methods across multiple benchmarks, and it can be paired with a planner and BEV controller for autonomous rollout and "what‑if" scenario generation.
The paper introduces a state‑aware framework that reconstructs both foreground and hidden scenes from single‑photon LiDAR histograms affected by partially transmissive occluders. It classifies each ray into no‑return, single‑return, or dual‑return states, guiding a two‑head neural field that jointly refines waveform reconstruction and geometry localization. A new paired dataset of occluded and clean LiDAR captures validates the method, showing improved depth and point‑cloud accuracy over existing baselines.