SpaRV: Sparse Radiance Voxels for Scalable Geometrically Consistent LiDAR-Visual Mapping and Rendering
We present SpaRV, a Sparse Radiance Voxel map rendered by differentiable ray marching, which lets irregular LiDAR scans supervise it alongside images.
ProofPaper ↗
Key points
- Robotic simulation, navigation, and inspection need photorealistic and geometrically accurate maps.
- Radiance fields achieve photorealism, but building them from robot data raises three challenges.
- Residual errors in camera poses and intrinsics blur the map.
- To bound GPU memory in training and rendering, SpaRV holds one submap at a time at full resolution and the rest at coarser octree levels.
Sources (1)
- [1]SpaRV: Sparse Radiance Voxels for Scalable Geometrically Consistent LiDAR-Visual Mapping and RenderingarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 02:12 AM
We present SpaRV, a Sparse Radiance Voxel map rendered by differentiable ray marching, which lets irregular LiDAR scans supervise it alongside images.
Robotic simulation, navigation, and inspection need photorealistic and geometrically accurate maps.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 19, 2026From Short Clips to Real-Time Streams: Identity-Aware Supervision and Context-Efficient Generation for Video Face Swapping
- Oct 18, 2026CusADi2: Latency-Optimized Batch Computation of Robot Kinematics, Jacobians, and Hessians on the GPU
- Oct 18, 2026GOFusion: Real-Time Gaussian-Occupancy Submap Fusion for Fisher-Informed Autonomous Robot Navigation
- Oct 18, 2026Google Open-Sources Harness2, a Recursive Agent Harnessing Framework
- Oct 18, 2026Elevatormusic/DeepSeek-V4.1-Flash-1017-EXL3-TensorFold-2x-DGX-Spark: DeepSeek-V4.1-Flash-1017 EXL3 2.9bpw on 2x NVIDIA DGX Spark (GB10) with TensorFold: 2.1-2.3x the vLLM kit's decode speed.
- Oct 16, 2026[Research] 2,788 tokens/s on a single RTX 5090 with a diffusion LLM — Introducing ORL (fully open source)