CRISP: Fixing Flying Pixels in Latent LiDAR Generation via Diffusion Decoding
cs.CV, cs.AI
Submitted: 2026-10-08
Updated: 2026-10-20
Comments: Accepted at NeurIPS 2026 (poster). 41 pages, 10 figures, 18 tables. Project page: https://andrea25512.github.io/CRISP/
Code: https://github.com/ThibaultGROUEIX/ChamferDistancePytorch
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
The gist: Latent LiDAR pipelines suffer from flying pixels: convolutional VAEs blur sharp radial depth discontinuities, yielding edge depths that back-project to points floating between surfaces.
Terminology
Abstract
Latent LiDAR pipelines suffer from flying pixels: convolutional VAEs blur sharp radial depth discontinuities, yielding edge depths that back-project to points floating between surfaces. We identify this as a major, directly correctable decoder bottleneck and introduce CRISP: a pixel-space diffusion decoder with a backbone-agnostic latent adapter, DiT-based denoiser, and support mask predictor. CRISP replaces video-VAE and LiDAR-native decoders alike while keeping the encoder and latent generator fixed. Across KITTI-360, SemanticKITTI, and nuScenes, replacing only the decoder reduces FSVD/FPVD by 50.5% on average across frozen backbones; for generic video VAEs, the reductions reach 71%/74%. On the LiDAR-native LiDM backbone, FRID drops by 71%, with the largest gains at depth discontinuities. In a pretrained LiDM world model, the same zero-shot replacement improves FSVD by 15.5%, narrowing the sim-to-real gap.
Sources
- Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets
- A Survey of World Models for Autonomous Driving
- Taming Transformers for Realistic Lidar Point Cloud Generation
- Gaussian Error Linear Units (GELUs)
- Auto-Encoding Variational Bayes
- LOGen: Toward Lidar Object Generation by Point Diffusion
- 3D and 4D World Modeling: A Survey
- WorldLens: Full-Spectrum Evaluations of Driving World Models in Real World
- L3DR: 3D-aware LiDAR Diffusion and Rectification
- PiD: Fast and High-Resolution Latent Decoding with Pixel Diffusion
- Physically Based Neural LiDAR Resimulation
- Adverse Weather Conditions Augmentation of LiDAR Scenes with Latent Diffusion Models
- Progressive Distillation for Fast Sampling of Diffusion Models
- R3DPA: Leveraging 3D Representation Alignment and RGB Pretrained Priors for LiDAR Scene Generation
- Score-Based Generative Modeling through Stochastic Differential Equations
- Range-Edit: Semantic Mask Guided Outdoor LiDAR Scene Editing
- Wan: Open and Advanced Large-Scale Video Generative Models
- Forging Spatial Intelligence: A Roadmap of Multi-Modal Data Pre-Training for Autonomous Systems
- Pixel-Perfect Depth with Semantics-Prompted Diffusion Transformers
- U4D: Uncertainty-Aware 4D World Modeling from LiDAR Sequences
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models