<-Complexity Reduction Go to ToC
(Informative)
1. Introduction
This chapter provides the Neural Network weights obtained by applying the process specified in:
- Design Procedure to two important cases.
- Complexity Reduction to the Neural Network of point 1.
2. Weights for important applications
The weights of the SD (540×960) to HD (1080×1920) and HD (1080×1920) to UHD (2160×3840) up-sampling filters will be downloadable from the MPAI Git.
The performance of the un-pruned and pruned network has been verified according to Table 1:
Table 1 –
Standard sequences | CatRobot, FoodMarket4, ParkRunning3. |
Bits/sample | 8 and 10 bit-depth per component. |
Colour space | YCbCr with 4:2:0 sub sampling. |
Encoding technologies | AVC, HEVC, and VVC. |
Encoding settings | Random Access and Low Delay at QPs 22, 27, 32, 37, 42, 47. |
Up-sampling | SD to HD and HD to UHD. |
Metrics | BD-Rate, BD-PSNR and BD-VMAF |
Deep-learning structure | Same for all QPs |
3. Weights of the complexity-reduced network
The number of parameters of the pruned network is about of the original
The loss in performance of the pruned network is less than 1% in BD-rate compared to the network out of the Design Procedure.
The performance of the un-pruned and pruned network has been verified using the same parameters of Table 1
Results show an impressive improvement for all coding technologies, and encoding options for all three objective metrics when compared with the currently used traditional bicubic interpolation. In Table 2, LD stands for low delay and RA for Random Access
Table 2 – Performance of the EVC-UFV Up-sampling Filter
HEVC (LD) | VVC (LD) | HEVC (RA) | VVC (RA) | |
SD to HD (using own trained filter) | 12.2% | 13.8% | 17.3% | 22.5% |
HD to UHD (using own trained filter) | 6% | 6.5% | 6.0% | 7.9% |
SD to HD (using HD to UHD filter) | 11.6% | 11.4% | 15.3% | 19.9% |
The parameters of the u-pruned and pruned networks will be available at the MPAI Git.
<-Complexity Reduction Go to ToC