1. Introduction

This chapter specifies:

  1. The neural network model (section 2).
  2. The weights reflecting the architecture of the neural network model obtained by
    1. Applying the process specified in Design Procedure (Section 3).
    2. Deriving a single filter applicable to the cases below (Section4):
      1. SD (540×960) to HD (1080×1920) up-sampling
      2. HD (1080×1920) to 4K (2160×3840) up-sampling.
    3. Applying the process specified Complexity Reduction to the weights of 2.1 (Section 5).

2. Neural Network Model

To be specified

3. Weights by applying the Design Procedure

The weights of the SD to HD up-sampling filter can be downloaded.

The weights of the HD to 4D up-sampling filter can be downloaded.

Note that the weights have been watermarked to protect the Intellectual Property of the sets of weights.

4. Weights for combined SD to HD and HD to 4K

marginal performance loss

5. Weights after application of Complexity Reduction

For both 3. and 4.?