This is a collection of pretrained models for Real-ESRGAN's Compact architecture. There are 1x, 2x, and 4x models, as well as 1x and 2x "UltraCompact" and "SuperUltraCompact" models (think of these as the equivalent to ESRGAN "lite" models). By using these are pretrains for your models, you can ensure that your models are able to be interpolated with other Compact models that were trained from these. These pretrains are compatible with most existing compact models.

Compact Pretrain

DAT_2_X2

2xESRGAN

Made to restore screentones and remove compression artifacts in manga images with widths between 650 and 900 pixels (can be more than 900).

2x Manga Ora

To upscale manga including halftones, instead of trying to smooth them out.

Pretrained model: MangaScaleV2(not released)

MangaScaleV3

Pretrained: RRDB_PSNR_x4.pth

The original RRDB_PSNR_x4.pth model converted to 1x, 2x, 8x and 16x scales, intended to be used as pretrained models for new models at those scales. These are compatible with victor's 4xESRGAN.pth conversions

2x PSNR Pretrained Model

Official paper pretrain model

SRFormer: Permuted Self-Attention for Single Image Super-Resolution

Yupeng Zhou 1, Zhen Li 1, Chun-Le Guo 1, Song Bai 2, Ming-Ming Cheng 1, Qibin Hou 1

1TMCC, School of Computer Science, Nankai University

2ByteDance, Singapore

arXiv

The official PyTorch implementation of SRFormer: Permuted Self-Attention for Single Image Super-Resolution (arxiv). SRFormer achieves state-of-the-art performance in

    classical image SR
    lightweight image SR
    real-world image SR

The results can be found here.

    Abstract: In this paper, we introduce SRFormer, a simple yet effective Transformer-based model for single image super-resolution. We rethink the design of the popular shifted window self-attention, expose and analyze several characteristic issues of it, and present permuted self-attention (PSA). PSA strikes an appropriate balance between the channel and spatial information for self-attention, allowing each Transformer block to build pairwise correlations within large windows with even less computational burden. Our permuted self-attention is simple and can be easily applied to existing super-resolution networks based on Transformers. Without any bells and whistles, we show that our SRFormer achieves a 33.86dB PSNR score on the Urban100 dataset, which is 0.46dB higher than that of SwinIR but uses fewer parameters and computations. We hope our simple and effective approach can serve as a useful tool for future research in super-resolution model design. Our code is publicly available at https://github.com/HVision-NKU/SRFormer.

Contents

    Installation & Dataset
    Training
    Testing
    Results
    Pretrain Models
    Citations
    License and Acknowledgement

Installation & Dataset

    python 3.8
    pyTorch >= 1.7.0

cd SRFormer
pip install -r requirements.txt
python setup.py develop

Dataset

We use the same training and testing sets as SwinIR, the following datasets need to be downloaded for training.
Task 	Training Set 	Testing Set
classical image SR 	DIV2K (800 training images) or DIV2K +Flickr2K (2650 images) 	Set5 + Set14 + BSD100 + Urban100 + Manga109
lightweight image SR 	DIV2K (800 training images) 	Set5 + Set14 + BSD100 + Urban100 + Manga109
real-world image SR 	DIV2K (800 training images) +Flickr2K (2650 images) + OST (10324 images for sky,water,grass,mountain,building,plant,animal) 	RealSRSet+5images
Training

    Please download the dataset corresponding to the task and place them in the folder specified by the training option in folder /options/train/SRFormer
    Follow the instructions below to train our SRFormer.

# train SRFormer for classical SR task
./scripts/dist_train.sh 4 options/train/SRFormer/train_SRFormer_SRx2_scratch.yml
./scripts/dist_train.sh 4 options/train/SRFormer/train_SRFormer_SRx3_scratch.yml
./scripts/dist_train.sh 4 options/train/SRFormer/train_SRFormer_SRx4_scratch.yml
# train SRFormer for lightweight SR task
./scripts/dist_train.sh 4 options/train/SRFormer/train_SRFormer_light_SRx2_scratch.yml
./scripts/dist_train.sh 4 options/train/SRFormer/train_SRFormer_light_SRx3_scratch.yml
./scripts/dist_train.sh 4 options/train/SRFormer/train_SRFormer_light_SRx4_scratch.yml

Testing

# test SRFormer for classical SR task
python basicsr/test.py -opt options/test/SRFormer/test_SRFormer_DF2Ksrx2.yml
python basicsr/test.py -opt options/test/SRFormer/test_SRFormer_DF2Ksrx3.yml
python basicsr/test.py -opt options/test/SRFormer/test_SRFormer_DF2Ksrx4.yml
# test SRFormer for lightweight SR task
python basicsr/test.py -opt options/test/SRFormer/test_SRFormer_light_DIV2Ksrx2.yml
python basicsr/test.py -opt options/test/SRFormer/test_SRFormer_light_DIV2Ksrx3.yml
python basicsr/test.py -opt options/test/SRFormer/test_SRFormer_light_DIV2Ksrx4.yml

Results

We provide the results on classical image SR, lightweight image SR, realworld image SR. More results can be found in the paper. The visual results of SRFormer will upload to google drive soon.

Classical image SR

    Results of Table 4 in the paper

    Results of Figure 4 in the paper

Lightweight image SR

    Results of Table 5 in the paper

    Results of Figure 5 in the paper

Model size comparison

    Results of Table 1 and Table 2 in the Supplementary Material

Realworld image SR

    Results of Figure 8 in the paper

Pretrain Models

Pretrain Models can be download from google drive. To reproduce the results in the article, you can download them and put them in the /PretrainModel folder.
Citations

You may want to cite:

@article{zhou2023srformer,
  title={SRFormer: Permuted Self-Attention for Single Image Super-Resolution},
  author={Zhou, Yupeng and Li, Zhen and Guo, Chun-Le and Bai, Song and Cheng, Ming-Ming and Hou, Qibin},
  journal={arXiv preprint arXiv:2303.09735},
  year={2023}
}

License and Acknowledgement

This project is released under the Apache 2.0 license. The codes are based on BasicSR, Swin Transformer, and SwinIR. Please also follow their licenses. Thanks for their awesome works.

SRFormerLight_SRx2_DIV2K

SuperUltraCompact Pretrain

[Swift-SRGAN - Rethinking Super-Resolution for real-time inference](https://github.com/Koushik0901/Swift-SRGAN)

In recent years, there have been several advancements in the task of image super-resolution using the state of the art Deep Learning-based architectures. Many super-resolution-based techniques previously published, require high-end and top-of-the-line Graphics Processing Unit (GPUs) to perform image super-resolution. With the increasing advancements in Deep Learning approaches, neural networks have become more and more compute hungry. We took a step back and, focused on creating a real-time efficient solution. We present an architecture that is faster and smaller in terms of its memory footprint. The proposed architecture uses Depth-wise Separable Convolutions to extract features and, it performs on-par with other super-resolution GANs (Generative Adversarial Networks) while maintaining real-time inference and a low memory footprint. A real-time super-resolution enables streaming high resolution media content even under poor bandwidth conditions. While maintaining an efficient trade-off between the accuracy and latency, we are able to produce a comparable performance model which is one-eighth (1/8) the size of super-resolution GANs and computes 74 times faster than super-resolution GANs.

NOTE: The author used the wrong file extensions for the models *on GitHub*. You will download a `.pth.tar` file. This is not actually a TAR file. Change the file extension to just `.pth` and the model will work.

Swift-SRGAN 2x

UltraCompact Pretrain

This is just a very basic 2x model to start your anime models with on SPAN, nothing special. I tried to keep it as basic as possible

2x-span_anime_pretrain

Official 2x pretrain for [SPAN](https://github.com/hongyuanyu/span).

spanx2_ch48

Purpose: pre-workout to speed up your workouts

While experimenting with color models for cugan, I just ran into problems that the official models are not imported correctly and have additional processing in addition to resizing. As a result, for my project I trained 2 x cugans for neosr; only resizing losses were applied to the dataset `(“lanczos”, “bicubic”, “bilenear”, “HAMMING”, “NEAREST”)`

wtp manga cover pretrained

Purpose: remove paper texture and screentone from independent manga scans

Well, I don’t even know how to describe it, a small breakthrough in my war with self-scans, I’m still not very happy with it, but maybe the lump will come in handy one day https://discord.com/channels/547949405949657098/547949806761410560/1204387665798103050

DWTP DP dat2 beta

Purpose: remove paper texture and screentone from independent manga scans

This is my first model for this, it can be said to remove poorly https://discord.com/channels/547949405949657098/1159598245417332796/1185562840304721921

DWTP DS dat2 v1

I've been experimenting a lot since the last descripton was released to save even more detail, and to get rid of some of the beginner's soreness. All three models are not yet perfect and I will develop them, but what is the essence of the three models, the models on omni is just a division of VL eats less large screenton, VH is just a re-trained model on a modified dataset and with VL as a prevoritolnoy shadowing that it would demolish more screenton, while not going to demolish all the textures and very large screenton. Esrgan model was trained on a dataset VH for use in the colab, it works in some places worse than Omni, so could not fight with overlapping screenton.

DWTP_descreenton_H_esrgan

DWTP_descreenton_VH4

DWTP_descreenton_VL4

Purpose: manga screentone removal

I haven’t shared models for a long time, I think one of these days I’ll upload a dat version and a less accurate esrgan. This version of the model was meant to be an attempt to get as much detail out of the scans as possible while removing the screentone, so some things had to be sacrificed, but overall it is better than all my previous models combined in terms of accuracy and tone.

[**examples**](https://slow.pics/c/zbOGXoDv)
[colab which contains the rest of my models](https://colab.research.google.com/drive/1w8fgsmh9fcHDtpQnYy77aKgpHEWlKpj-?usp=sharing#scrollTo=aYbqM1pWkPTT)
||In general, this is a collab for convenient scaling of manga in the ru segment of translators and there are other people’s models there, if I suddenly added someone else’s model that has a license, please let me know, I’ll correct it||

DWTP ds esrgan 5

I just decided to finish off the previous one, made the manga design more clear and corrected the deviation in color temperature by adding a more diverse set of colors

WTP UDS Esrgan

Purpose: Grayscale pencil/sketch style art with paper textures (Made in Abyss specific)

Old model trained for pencil sketch style art, good at restoring the paper and pencil and other textures , specifically the Made in Abyss manga, but should work for art with a similar style.

Comparison:
https://slow.pics/c/D0MuvwLY

eula digimanga MiA 65k

Black and white digital manga with halftones.

eula digimanga bw v1

Vast improvement over v1 in low frequency detail; moiré and artifacting reduced significantly and less random noise from JPEG artefacts in the input. Also now only works on 1 channel images, so it runs slightly faster on average and resulting images are much smaller but it might not work on some ESRGAN implementations, I personally recommend using chaiNNer. v1 may still be better in some edge cases.

There's also a supplementary 1x model that denoises very low quality LRs and smooths halftones so the image works better with the 4x model. Only trained it to help build the dataset and it's useless for already decent-ish LRs but may help you in some situations.

eula digimanga bw v2 nc1

In general, what was the point? The project was conceived for a manga scale with a screentone using the method of superimposing a new screentone, but for this the old screentone must be removed, but this is not all the same important contour. With the screentone overlay method, the contour should be as close to 1 bit as possible, otherwise we will see full contours, wavy contours, etc. This is the whole idealogy of this model.

wtp_descreentone

Purpose: pre-workout to speed up your workouts

I see that many people are trying manga scale models on the Omni, and to speed up the process I decided to do preliminary training, the only loss is the resize: (“lanczos”, “bicubic”, “bilenear”, “HAMMING”)

Original link: https://drive.google.com/drive/folders/1ln1ZneTvqPbvi8nsJwzjMoYSDQ_DcHBS?usp=sharing

Architecture	Real-CUGAN
Scale	2x
Color Mode	RGB
License	None
Date	2024-02-19
Dataset	custom dataset of manga covers and color manga in equal quality with pixiv
Dataset size	110730
Training iterations	312000
Training epochs	14
Training batch size	6
Training HR size	256
Training OTF	No

wtp manga cover pretrained

Similar Models