CurveCodec 2: Skeleton-agnostic animation compression with a learned entropy model

2026-10-03

AI Summary

This study addresses the limited generalizability and excessive worst-case errors of existing skeletal animation compression methods by proposing a high-fidelity, bit-level compression framework applicable to arbitrary skeletons. Methodologically, we identify that redundancy primarily stems from curve prediction rather than interpolation. Accordingly, we introduce logarithmic mapping and closed-loop quantization, employ rate-distortion optimization for keyframe selection, and design a bit-exact learned entropy model for integer inference coding to ensure cross-platform decoding consistency. This work is the first to simultaneously satisfy both average and worst-case error constraints defined by the ACL benchmark. At default precision, the proposed method reduces the bitrate to 0.37× that of ACL while supporting real-time single-core CPU decoding and exhibiting zero-shot generalization capabilities.

Abstract

Skeletal motion is stored as every joint's transform at every frame, yet most of it is implied by the body rather than by what the motion is about. Compression is one way to ask what a motion must still say once the body is known, and a production codec must answer it for any skeleton with a stated error bound. Our earlier codec, CurveCodec, matched the mean error of ACL, the production library of modern game engines, with a learned prior over sparse anchors, but not ACL's worst case, and it counted its payload as floats rather than bits. Here we ask where the redundancy of skeletal motion lies and which part of a codec a learned model should take over. Measurements give three answers. At production precision the largest saving comes from predicting each quantized curve from its own past, the second from choosing per joint, in closed loop through the hierarchy, which samples not to code. On the gaps such an encoder leaves, a nearest-neighbour oracle over millions of training samples is no better than linear interpolation, and no learned in-betweener we tried paid for itself. What a network does learn is the distribution of the residuals the codec must send. CurveCodec 2 codes every sub-track as a curve in the log map, quantized in closed loop and thinned to rate-distortion-selected keys, with residuals entropy-coded under a small learned model whose integer inference is bit-exact across platforms. Two contracts are verified on every decoded clip: ACL's own worst case per joint within a stated tolerance, or ACL's mean error per clip. On a held-out test side of 4,472 clips from 33 datasets, CurveCodec 2 needs 0.37x ACL's bytes at ACL's default precision of 0.01 cm under the worst-case contract and 0.22x at 0.1 cm under the mean contract, decodes on one CPU core, and transfers without retraining to a species absent from training. Project page: https://rubbly.cn/publications/curvecodec/

Problem

• animation compression
• skeletal motion
• entropy coding
• rate-distortion
• skeleton-agnostic

Innovation

• Skeleton-agnostic animation compression
• Learned entropy model
• Log map curve quantization
• Rate-distortion keyframe selection
• Bit-exact integer inference