Skip to content

Latest commit

 

History

History
44 lines (29 loc) · 2.53 KB

cmu_performance.md

File metadata and controls

44 lines (29 loc) · 2.53 KB

Carnegie-Mellon University database performance

v1.3.0 v1.2.0 v1.1.0
Compressed size 71.00 MB 71.01 MB 67.06 MB
Compression ratio 20.13 : 1 20.13 : 1 21.31 : 1
Compression time 2m 22.3s 6m 9.71s 35m 19.96s
Compression speed 10285.52 KB/sec 3958.99 KB/sec 690.43 KB/sec
Max error 0.0725 cm 0.0725 cm 0.0725 cm
Error 99th percentile 0.0089 cm 0.0089 cm 0.0090 cm
Samples below error threshold 99.86 % 99.86 % 99.83 %

Note: Starting with v1.3.0 measurements have been made with a Ryzen 2950X CPU while prior versions used an Intel i7 6850K.

Data and method used

To compile the statistics, the animation database from Carnegie-Mellon University is used. The raw animation clips in FBX form can be found on the Unity asset store here. They were converted to the ACL file format using the fbx2acl script. Data available upon request, it is far too large for GitHub.

  • Number of clips: 2534
  • Sample rate: 24 FPS
  • Total duration: 9h 49m 37.58s
  • Raw size: 1429.38 MB (10x float32 * num bones * num samples)

ACL supports various compression methods but only the overall best variant will be tracked here (see get_default_compression_settings() for details).

The error is measured 3cm away from each bone to simulate the visual mesh skinning process as described here with the default error threshold of 0.01cm. The 99th percentile and the number of samples below the error threshold are calculated by measuring the error on every bone at every sample.

The performance of ACL in Unreal Engine 4 is tracked by the plugin here.

Results in images

Compression ratio VS max error per clip

Compression ratio distribution

Max error distribution

Distribution of the error for every bone at every key frame

Distribution of the error for every bone at every key frame (top 10%)

Distribution of clip durations