Squeeze3D: Your 3D Generation Model is Secretly an Extreme Neural Compressor

2025-06-09

AI Summary

This work addresses the challenge of neural compression for cross-modal 3D data—specifically meshes, point clouds, and radiance fields—under extremely high compression ratios, where implicit priors from pretrained 3D generative models (e.g., NeRFs or 3D diffusion models) are difficult to leverage. We propose Squeeze3D, a framework that introduces a learnable latent-space mapping network to bridge a generic 3D encoder and a frozen pretrained generative model, eliminating the need for fine-tuning on real 3D data. To our knowledge, this is the first demonstration that pretrained 3D generative models can serve as efficient, universal neural compressors—trained solely on synthetic data and supporting unified compression across modalities without object-level customization. Experiments show compression ratios of 2187× (textured meshes), 55× (point clouds), and 619× (radiance fields), with reconstruction quality competitive with state-of-the-art methods and significantly reduced encoding/decoding latency.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: 3D Computer Vision
• Data Mining & Knowledge Management: Data Compression

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

We propose Squeeze3D, a novel framework that leverages implicit prior knowledge learnt by existing pre-trained 3D generative models to compress 3D data at extremely high compression ratios. Our approach bridges the latent spaces between a pre-trained encoder and a pre-trained generation model through trainable mapping networks. Any 3D model represented as a mesh, point cloud, or a radiance field is first encoded by the pre-trained encoder and then transformed (i.e. compressed) into a highly compact latent code. This latent code can effectively be used as an extremely compressed representation of the mesh or point cloud. A mapping network transforms the compressed latent code into the latent space of a powerful generative model, which is then conditioned to recreate the original 3D model (i.e. decompression). Squeeze3D is trained entirely on generated synthetic data and does not require any 3D datasets. The Squeeze3D architecture can be flexibly used with existing pre-trained 3D encoders and existing generative models. It can flexibly support different formats, including meshes, point clouds, and radiance fields. Our experiments demonstrate that Squeeze3D achieves compression ratios of up to 2187x for textured meshes, 55x for point clouds, and 619x for radiance fields while maintaining visual quality comparable to many existing methods. Squeeze3D only incurs a small compression and decompression latency since it does not involve training object-specific networks to compress an object.

Problem

• Compress 3D data at extreme ratios using generative models
• Bridge latent spaces between encoder and generative models
• Support multiple 3D formats without object-specific training

Innovation

• Leverages pre-trained 3D generative models
• Uses trainable mapping networks
• Supports multiple 3D data formats