Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,65 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: cc-by-nc-sa-4.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-nc-sa-4.0
|
| 3 |
+
pipeline_tag: image-to-image
|
| 4 |
+
tags:
|
| 5 |
+
- pytorch
|
| 6 |
+
- super-resolution
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
## 2x-AnimeSharpV4 & Fast
|
| 10 |
+
|
| 11 |
+
**Scale:** 2
|
| 12 |
+
|
| 13 |
+
**Architecture:** RCAN & RCAN PixelUnshuffle
|
| 14 |
+
|
| 15 |
+
**Links:** [Github Release](<https://github.com/Kim2091/Kim2091-Models/releases/tag/2x-AnimeSharpV4>)
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
**Author:** Kim2091
|
| 19 |
+
|
| 20 |
+
**License:** CC BY-NC-SA 4.0
|
| 21 |
+
|
| 22 |
+
**Purpose:** Anime
|
| 23 |
+
|
| 24 |
+
**Subject:**
|
| 25 |
+
|
| 26 |
+
**Input Type:** Images
|
| 27 |
+
|
| 28 |
+
**Date:** 1-7-25
|
| 29 |
+
|
| 30 |
+
**Size:**
|
| 31 |
+
|
| 32 |
+
**I/O Channels:** 3(RGB)->3(RGB)
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
**Dataset:** ModernAnimation1080_v3 & digital_art_v3
|
| 36 |
+
|
| 37 |
+
**Dataset Size:** 6k & 20k
|
| 38 |
+
|
| 39 |
+
**OTF (on the fly augmentations):** No
|
| 40 |
+
|
| 41 |
+
**Pretrained Model:** 2x-AnimeSharpV3_RCAN & database's 12k PU checkpoint
|
| 42 |
+
|
| 43 |
+
**Iterations:** 100k RCAN & 400k RCAN PU
|
| 44 |
+
|
| 45 |
+
**Batch Size:** 8
|
| 46 |
+
|
| 47 |
+
**GT Size:** 64
|
| 48 |
+
|
| 49 |
+
|
| 50 |
+
**Description:** This is a successor to AnimeSharpV3 based on RCAN instead of ESRGAN. It outperforms both versions of AnimeSharpV3 in every capacity. It's sharper, retains *even more* detail, and has very few artifacts. It is __extremely faithful__ to the input image, even with heavily compressed inputs.
|
| 51 |
+
|
| 52 |
+
|
| 53 |
+
Currently it is __NOT compatible with chaiNNer__, but will be available on the nightly build soon (hopefully).
|
| 54 |
+
|
| 55 |
+
|
| 56 |
+
The `2x-AnimeSharpV4_Fast_RCAN_PU` model is trained on RCAN PixelUnshuffle. This is much faster, but comes at the cost of quality. I believe the model is ~95% the quality of the full V4 RCAN model, but ~6x faster in Pytorch and ~4x faster in TensorRT. This model is ideal for video processing, and as such was trained to handle MPEG2 & H264 compression.
|
| 57 |
+
|
| 58 |
+
__Comparisons:__
|
| 59 |
+
|
| 60 |
+
https://slow.pics/c/63Qu8HTN
|
| 61 |
+
|
| 62 |
+
https://slow.pics/c/DBJPDJM9
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
|