# Pointsoup: High-Performance and Extremely Low-Decoding-Latency Learned Geometry Codec for Large-Scale Point Cloud Scenes

arXiv: https://arxiv.org/abs/2404.13550 · HTML: https://mlsanity.com/paper/2404.13550/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 33 references that are in the map.
- First posted on arXiv 21 Apr 2024.

## Builds on (most cited first, 17 of 17)

- [ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes](https://mlsanity.com/paper/1702.04405/) (2017)
- [ShapeNet: An Information-Rich 3D Model Repository](https://mlsanity.com/paper/1512.03012/) (2015)
- [Dynamic Graph CNN for Learning on Point Clouds](https://mlsanity.com/paper/1801.07829/) (2018)
- [Point Transformer](https://mlsanity.com/paper/2011.00931/) (2020)
- [PCT: Point cloud transformer](https://mlsanity.com/paper/2012.09688/) (2020)
- [End-to-end Optimized Image Compression](https://mlsanity.com/paper/1611.01704/) (2016)
- [Point Transformer V2: Grouped Vector Attention and Partition-based Pooling](https://mlsanity.com/paper/2210.05666/) (2022)
- [Multiscale Point Cloud Geometry Compression](https://mlsanity.com/paper/2011.03799/) (2020)
- [Sparse Tensor-based Multiscale Representation for Point Cloud Geometry Compression](https://mlsanity.com/paper/2111.10633/) (2021)
- [OctAttention: Octree-Based Large-Scale Contexts Model for Point Cloud Compression](https://mlsanity.com/paper/2202.06028/) (2022)
- [ProxyFormer: Proxy Alignment Assisted Point Cloud Completion with Missing Part Sensitive Transformer](https://mlsanity.com/paper/2302.14435/) (2023)
- [Density-preserving Deep Point Cloud Compression](https://mlsanity.com/paper/2204.12684/) (2022)
- [Patch-Based Deep Autoencoder for Point Cloud Geometry Compression](https://mlsanity.com/paper/2110.09109/) (2021)
- [Learning-Driven Lossy Image Compression; A Comprehensive Survey](https://mlsanity.com/paper/2201.09240/) (2022)
- [Multiscale Latent-Guided Entropy Model for LiDAR Point Cloud Compression](https://mlsanity.com/paper/2209.12512/) (2022)
- [IPDAE: Improved Patch-Based Deep Autoencoder for Lossy Point Cloud Geometry Compression](https://mlsanity.com/paper/2208.02519/) (2022)
- [Dynamic Local Feature Aggregation for Learning on Point Clouds](https://mlsanity.com/paper/2301.02836/) (2023)

## Built on by (most cited first, 5 of 5)

- [RENO: Real-Time Neural Compression for 3D LiDAR Point Clouds](https://mlsanity.com/paper/2503.12382/) (2025)
- [UniPCGC: Towards Practical Point Cloud Geometry Compression via an Efficient Unified Approach](https://mlsanity.com/paper/2503.18541/) (2025)
- [Diff-PCC: Diffusion-based Neural Compression for 3D Point Clouds](https://mlsanity.com/paper/2408.10543/) (2024)
- [AnyPcc: Compressing Any Point Cloud with a Single Universal Model](https://mlsanity.com/paper/2510.20331/) (2025)
- [FLaTEC: Frequency-Disentangled Latent Triplanes for Efficient Compression of LiDAR Point Clouds](https://mlsanity.com/paper/2511.20065/) (2025)

## Abstract (arXiv)

Despite considerable progress being achieved in point cloud geometry compression, there still remains a challenge in effectively compressing large-scale scenes with sparse surfaces. Another key challenge lies in reducing decoding latency, a crucial requirement in real-world application. In this paper, we propose Pointsoup, an efficient learning-based geometry codec that attains high-performance and extremely low-decoding-latency simultaneously. Inspired by conventional Trisoup codec, a point model-based strategy is devised to characterize local surfaces. Specifically, skin features are embedded from local windows via an attention-based encoder, and dilated windows are introduced as cross-scale priors to infer the distribution of quantized features in parallel. During decoding, features undergo fast refinement, followed by a folding-based point generator that reconstructs point coordinates with fairly fast speed. Experiments show that Pointsoup achieves state-of-the-art performance on multiple benchmarks with significantly lower decoding complexity, i.e., up to 90~160× faster than the G-PCCv23 Trisoup decoder on a comparatively low-end platform (e.g., one RTX 2080Ti). Furthermore, it offers variable-rate control with a single neural model (2.9MB), which is attractive for industrial practitioners.