CurveCodec 2: Skeleton-agnostic animation compression with a learned entropy model
CurveCodec 2: Skeleton-agnostic animation compression with a learned entropy model
This research paper introduces CurveCodec 2, an animation compression system that achieves significant data reduction for skeletal motion. The method utilizes a learned entropy model to code residuals after quantizing curves in the log map and thinning them based on rate-distortion criteria. On a held-out test set of 4,472 clips from 33 datasets, CurveCodec 2 requires only 0.37x the bytes of the industry-standard ACL library at its default precision (0.01 cm) under a worst-case error contract, while decoding on a single CPU core and transferring without retraining to unseen skeleton types.
The authors present CurveCodec 2, a skeleton-agnostic animation compression method that achieves 0.37x the byte size of the industry-standard ACL library under worst-case error contracts on a held-out test set of 4,472 clips from 33 datasets.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
bytes
0.37
View original evidence
odec 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/
Open source S6bytes
0.22
View original evidence
odec 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/
Open source S6clips
4,472
Reported trials
View original evidence
ub-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, CurveC
Open source S5datasets
33
Reported trials
View original evidence
ub-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, CurveC
Open source S5percent
0.01
View original evidence
odec 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/
Open source S6percent
0.1
View original evidence
odec 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/
Open source S6core
1
View original evidence
odec 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/
Open source S6
- paper: not_reported
- code: Reported pending Open source S6
- weights: not_reported
- dataset: not_reported
- code: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
odec 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/
Open source S6
ub-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, CurveC
Open source S5
rror 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 f
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org