Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,41 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: cc-by-4.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
[Link to Github Release](https://github.com/Phhofm/models/releases/tag/4xNomos8k_atd_jpg)
|
| 6 |
+
|
| 7 |
+
|
| 8 |
+
# 4xNomos8k_atd_jpg
|
| 9 |
+
|
| 10 |
+
Name: 4xNomos8k_atd_jpg
|
| 11 |
+
License: CC BY 4.0
|
| 12 |
+
Author: Philip Hofmann
|
| 13 |
+
Network: [ATD](https://github.com/LabShuHangGU/Adaptive-Token-Dictionary)
|
| 14 |
+
Scale: 4
|
| 15 |
+
Release Date: 22.03.2024
|
| 16 |
+
Purpose: 4x photo upscaler, handles jpg compression
|
| 17 |
+
Iterations: 240'000
|
| 18 |
+
epoch: 152
|
| 19 |
+
batch_size: 6, 3
|
| 20 |
+
HR_size: 128, 192
|
| 21 |
+
Dataset: nomos8k
|
| 22 |
+
Number of train images: 8492
|
| 23 |
+
OTF Training: Yes
|
| 24 |
+
Pretrained_Model_G: 003_ATD_SRx4_finetune
|
| 25 |
+
|
| 26 |
+
Description:
|
| 27 |
+
4x photo upscaler which handles jpg compression. This model will preserve noise. Trained on the very recently released (~2 weeks ago) Adaptive-Token-Dictionary network.
|
| 28 |
+
|
| 29 |
+
Training details:
|
| 30 |
+
AdamW optimizer with U-Net SN discriminator and BFloat16.
|
| 31 |
+
Degraded with otf jpg compression down to 40, re-compression down to 40, together with resizes and the blur kernels.
|
| 32 |
+
Losses: PixelLoss using CHC (Clipped Huber with Cosine Similarity Loss), PerceptualLoss using Huber, GANLoss, [LDL](https://github.com/csjliang/LDL) using Huber, YCbCr Color Loss (bt601) and Luma Loss (CIE XYZ) on [neosr](https://github.com/muslll/neosr).
|
| 33 |
+
|
| 34 |
+
7 Examples:
|
| 35 |
+
[Slowpics](https://slow.pics/s/uwnoI435)
|
| 36 |
+
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
|