← Back to all news
arXiv Computer Vision

Extrapolation Matters: Sparse-Only Depth Completion Beyond LiDAR Coverage

The paper introduces a depth completion method that focuses on extrapolating depth beyond LiDAR coverage, a scenario often overlooked in current research. It uses a frozen monocular depth foundation model, aligns it with sparse LiDAR points, and refines the result with a module that fuses the prior and sparse anchors to produce dense metric depth. A novel high‑frequency normal‑consistency loss and a new evaluation protocol that separates LiDAR‑supported and beyond‑coverage regions are also presented, demonstrating strong accuracy in unobserved areas while maintaining performance where LiDAR data exists.

Related stories

arXiv Computer Vision
Sep 25

FounRef: Robust, Structure-Preserving, and Fast Metric Refinement of Frozen Monocular Foundation Priors with Sparse Anchors

FounRef is a training‑free method that refines frozen monocular foundation priors into dense metric depth by aligning them with sparse metric anchors. It validates anchors against the prior’s predictions, rejects misaligned ones, and applies a structure‑preserving solver to correct depth globally and locally while preserving fine geometry. The approach works out of the box on unseen cameras and scenes, achieving up to 24% lower depth error, 92% lower surface‑normal noise, and nearly 15× faster inference than a leading depth‑completion network.

By Dan Halperin, Mirko M\"ahlisch
benchmarkssafety
More like this →
arXiv Computer Vision
2d ago

eSharp: Diffusion-based Sharp Monocular Metric Event-to-Depth Estimation

eSharp is a monocular depth estimation framework that uses an event camera and pretrained diffusion priors to produce accurate, sharp metric depth from sparse LiDAR ground truth. It tackles challenges of fast motion and high-dynamic-range scenes by cross‑modal distillation, generating dense pseudo‑labels from a co‑registered frame camera. Experiments demonstrate that eSharp surpasses existing event‑to‑depth baselines in both depth accuracy and spatial fidelity, positioning diffusion models as a strong foundation for event‑driven 3D perception.

By Qiushi Dai, Aniket Dashpute, Yuhao Liu, Rishubh Parihar, Guha Balakrishnan, Ashok Veeraraghavan
agentsdiffusioncomputer-visionefficiencymultimodal
More like this →
Hugging Face Trending Papers
Sep 8

DXPR: Depth-Based Vision-LiDAR Cross-Modal Place Recognition Using Vision Foundation Models

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.

agentsroboticsmultimodalsafety
More like this →
Hugging Face Trending Papers
Jun 1

Honey, I Shrunk the Arc de Triomphe!

Metric scale monocular geometry estimation has seen significant progress through large-scale data aggregation, yet current foundation models suffer from a persistent ''scale-collapse'' phenomenon: distant landmarks and vast landscapes are metrically underestimated. We hypothesize that this performance gap stems from a training data bottleneck, where existing metric-scale datasets are hardware-constrained to homogenous vehicle-captured LiDAR or short-range indoor scans, or consist of synthetic data that lacks the semantic complexity of the physical world.

fine-tuningbenchmarks
More like this →
arXiv Computer Vision
Oct 5

LightLoc++: Sensor-Robust Representation Learning for Efficient Outdoor LiDAR Localization

LightLoc++ introduces a sensor‑robust framework for outdoor LiDAR localization that decouples a pretrained backbone from lightweight scene‑specific heads, enabling efficient new‑scene training. The authors present SULID, a synchronized urban multi‑LiDAR dataset featuring 32‑, 64‑, and 128‑beam sensors, which is used to pretrain a backbone via cross‑sensor consistency learning. Additional techniques such as sample classification guidance and redundant sample downsampling reduce regression ambiguity and computational load, achieving state‑of‑the‑art performance with minimal new‑scene training cost.

By Wen Li, Shangshu Yu, Dunqiang Liu, Shaoyang Chen, Qiming Xia, Sheng Ao, Siqi Shen, Chenglu Wen, Cheng Wang
benchmarks
More like this →