RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: You, Kang, Chen, Tong, Ding, Dandan, Asif, M. Salman, Ma, Zhan
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910877621420032
author You, Kang
Chen, Tong
Ding, Dandan
Asif, M. Salman
Ma, Zhan
author_facet You, Kang
Chen, Tong
Ding, Dandan
Asif, M. Salman
Ma, Zhan
contents Despite the substantial advancements demonstrated by learning-based neural models in the LiDAR Point Cloud Compression (LPCC) task, realizing real-time compression - an indispensable criterion for numerous industrial applications - remains a formidable challenge. This paper proposes RENO, the first real-time neural codec for 3D LiDAR point clouds, achieving superior performance with a lightweight model. RENO skips the octree construction and directly builds upon the multiscale sparse tensor representation. Instead of the multi-stage inferring, RENO devises sparse occupancy codes, which exploit cross-scale correlation and derive voxels' occupancy in a one-shot manner, greatly saving processing time. Experimental results demonstrate that the proposed RENO achieves real-time coding speed, 10 fps at 14-bit depth on a desktop platform (e.g., one RTX 3090 GPU) for both encoding and decoding processes, while providing 12.25% and 48.34% bit-rate savings compared to G-PCCv23 and Draco, respectively, at a similar quality. RENO model size is merely 1MB, making it attractive for practical applications. The source code is available at https://github.com/NJUVISION/RENO.
format Preprint
id arxiv_https___arxiv_org_abs_2503_12382
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds
You, Kang
Chen, Tong
Ding, Dandan
Asif, M. Salman
Ma, Zhan
Computer Vision and Pattern Recognition
Image and Video Processing
Despite the substantial advancements demonstrated by learning-based neural models in the LiDAR Point Cloud Compression (LPCC) task, realizing real-time compression - an indispensable criterion for numerous industrial applications - remains a formidable challenge. This paper proposes RENO, the first real-time neural codec for 3D LiDAR point clouds, achieving superior performance with a lightweight model. RENO skips the octree construction and directly builds upon the multiscale sparse tensor representation. Instead of the multi-stage inferring, RENO devises sparse occupancy codes, which exploit cross-scale correlation and derive voxels' occupancy in a one-shot manner, greatly saving processing time. Experimental results demonstrate that the proposed RENO achieves real-time coding speed, 10 fps at 14-bit depth on a desktop platform (e.g., one RTX 3090 GPU) for both encoding and decoding processes, while providing 12.25% and 48.34% bit-rate savings compared to G-PCCv23 and Draco, respectively, at a similar quality. RENO model size is merely 1MB, making it attractive for practical applications. The source code is available at https://github.com/NJUVISION/RENO.
title RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds
topic Computer Vision and Pattern Recognition
Image and Video Processing
url https://arxiv.org/abs/2503.12382