2x-AnimeSharpV4 & Fast
Scale: 2
Architecture: RCAN & RCAN PixelUnshuffle
Links: Github Release
Author: Kim2091
License: CC BY-NC-SA 4.0
Purpose: Anime
Subject:
Input Type: Images
Date: 1-7-25
Size:
I/O Channels: 3(RGB)->3(RGB)
Dataset: ModernAnimation1080_v3 & digital_art_v3
Dataset Size: 6k & 20k
OTF (on the fly augmentations): No
Pretrained Model: 2x-AnimeSharpV3_RCAN & database's 12k PU checkpoint
Iterations: 100k RCAN & 400k RCAN PU
Batch Size: 8
GT Size: 64
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.
Currently it is NOT compatible with chaiNNer, but will be available on the nightly build soon (hopefully).
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.
To use the Pytorch version of the model right now, you can manually update your version of the spandrel library in chaiNNer or another tool to this version: https://github.com/Kim2091/spandrel/actions/runs/12701005765
Comparisons:
