Skip to content

Latest commit

 

History

History
49 lines (36 loc) · 2.97 KB

paragon_performance.md

File metadata and controls

49 lines (36 loc) · 2.97 KB

Paragon database performance

ACL Plugin v2.1.0 ACL Plugin v2.0.0 UE v5.2.0
Compressed size 183.98 MB 224.30 MB 391.66 MB
Compression ratio 23.24 : 1 19.06 : 1 10.92 : 1
Compression time 19m 22.95s 8m 32.77s 33m 24.28s
Compression speed 3765.21 KB/sec 8539.37 KB/sec 2184.70 KB/sec
Max ACL error 1.2065 cm 0.8622 cm 1.2169 cm
Max UE error 0.8434 cm 0.8602 cm 0.8491 cm
ACL Error 99th percentile 0.0197 cm 0.0095 cm 0.0358 cm
Samples below ACL error threshold 94.78 % 99.13 % 82.81 %

ACL was smaller for 6469 clips (98.64 %)
ACL was more accurate for 4829 clips (73.64 %)
ACL has faster compression for 5302 clips (80.85 %)
ACL was smaller, better, and faster for 3896 clips (59.41 %)

Would the ACL Plugin have been included in the Automatic Compression permutations tried, it would have won for 6479 clips (98.80 %)

Note: Numbers for ACL 2.0 were extracted with UE 4.25.

Data and method used

To compile these statistics, a large number of animations from Paragon are used. In October 2017 the animations were manually extracted and converted to the ACL file format losslessly. The data is sadly NOT available upon request. Epic has permitted Nicholas Frechette to use them for research purposes only under a non-disclosure agreement.

Note: Epic has since released Paragon assets publicly in early 2018, once I get around to it, it will be extracted along with updated stats.

  • Number of clips: 6558
  • Total duration: 7h 0m 45.27s
  • Raw size: 4276.11 MB (10x float32 * num bones * num samples)

The data set contains among other things:

  • Lots of characters with varying number of bones
  • Animated objects of various shape and form
  • Very short and very long clips
  • Clips with unusual sample rate (as low as 2 FPS!)
  • World space clips
  • Lots of 3D scale
  • Lots of other exotic clips

To measure and extract the compression statistics, the provided commandlet is used along with a python script to parse the results.

The ACL Plugin uses the default settings with an error threshold of 0.01cm while UE uses the Automatic Compression with an error threshold (master tolerance) of 0.1 cm. Both error thresholds used are suitable for production use. The 99th percentile and the number of samples below the ACL error threshold are calculated by measuring the error with ACL on every bone at every sample.

ACL and UE both use separate methods to measure the error and both values are shown for transparency.