---
license: bsd-3-clause
pipeline_tag: image-to-image
library_name: onnx
tags:
- onnx
- moonbox
- real-esrgan
- super-resolution
- upscaling
- image-to-image
- esrgan
---
# Real-ESRGAN ONNX
ONNX builds of [Real-ESRGAN](https://github.com/xinntao/Real-ESRGAN), made by Moonbox for on-device use with ONNX Runtime (CUDA, DirectML, CPU). The
model is a single graph: each precision is one folder holding `model.onnx`.
## Files
The last column is the benchmark: each precision's output against the fp32 graph's on the same eval pictures, and
its own score on the eval set (details under *How to run*).
| Precision | File | Size | sha256 | Benchmark |
|---|---|---|---|---|
| fp16 | [`fp16/model.onnx`](https://huggingface.co/moonbox-dev/Real-ESRGAN-ONNX/blob/main/fp16/model.onnx) | 0.03 GB | `50eedabe29b4` | vs fp32 PSNR 59.1 / 59.0 dB (Set14 / BSD100), vs ground truth PSNR-Y 24.73 / 24.22, SSIM-Y 0.673 / 0.629; 171 ms a picture |
| int8 | [`int8/model.onnx`](https://huggingface.co/moonbox-dev/Real-ESRGAN-ONNX/blob/main/int8/model.onnx) | 0.02 GB | `f2c1fb815ce1` | vs fp32 PSNR 38.8 / 39.0 dB (worst 29.9), vs ground truth PSNR-Y 24.80 / 24.30, SSIM-Y 0.674 / 0.630; 92 ms a picture |
## How to run
Real-ESRGAN x4plus, the general 4x upscaler for real-world photos, exported to ONNX from the official
`RealESRGAN_x4plus.pth` (GitHub release v0.1.0). Height and width are free: the graph takes any picture size.
**ONNX Runtime** (RGB in 0-1, output 4x larger):
```python
import numpy as np, onnxruntime as ort
from PIL import Image
sess = ort.InferenceSession("int8/model.onnx", providers=["CUDAExecutionProvider", "CPUExecutionProvider"])
x = np.asarray(Image.open("photo.png").convert("RGB"), np.float32).transpose(2, 0, 1)[None] / 255
y = sess.run(None, {"input": x})[0][0]
Image.fromarray((np.clip(y.transpose(1, 2, 0), 0, 1) * 255).round().astype(np.uint8)).save("photo_x4.png")
```
Memory grows with the picture (the network runs at the output size in its last layers): for pictures above about
1000 x 1000, upscale in tiles (for example 256 px with 16 px of overlap) and stitch.
**Precisions.** fp16: weights and math in fp16, inputs and outputs fp32. int8: weights int8 per output channel
behind DequantizeLinear, activations in float (weight-only: no calibration; calibrated activation quantization
hurts GAN upscalers' fine detail). No int4 (the network has no MatMul layers) and no bf16 (no bf16 convolution in
ONNX Runtime's CUDA provider).
**Benchmark.** Set14 and BSD100 at 4x (bicubic-downscaled inputs, the standard SR test sets), ONNX Runtime 1.30 CUDA
provider on an RTX 3080 Laptop GPU. The last column gives each precision's PSNR against the fp32 graph's output and
its PSNR / SSIM (Y channel, 4 px border cropped) against the ground truth. Real-ESRGAN is trained for real-world
degradations rather than bicubic ones, so its ground-truth PSNR is lower than a bicubic-trained model's; the
numbers compare these precisions with each other. On this GPU the fp16 graph ran slower than fp32 (171 vs 86 ms a
picture at these sizes); fp16 halves the file and memory, int8 quarters the file at fp32 speed.
## How these were made
From [Real-ESRGAN](https://github.com/xinntao/Real-ESRGAN) at revision `a4abfb2979`, starting from an fp32 ONNX graph (where it came from is said above):
fp16 with onnxconverter-common (inputs and outputs kept fp32), int8 with ONNX Runtime static QDQ quantization
(per-channel int8 weights, uint8 activations, MinMax calibration on pictures outside the eval set), int4 with ONNX
Runtime's MatMulNBits (weight-only, block 32) where the graph is mostly matrix multiplications. bf16 is left out for
convolutional graphs: ONNX Runtime's CUDA provider has no bf16 convolution. Every graph was benchmarked before
upload; the `sha256` column above is what the files must hash to.
## License
BSD-3-Clause. These files are a derivative of the original model and carry its license: it is copied in this
repository, and it governs your use of these files. The original model card follows unchanged.
---
# Original model card
##
š[**Demos**](#-demos-videos) **|** š©[**Updates**](#-updates) **|** ā”[**Usage**](#-quick-inference) **|** š°[**Model Zoo**](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/model_zoo.md) **|** š§[Install](#-dependencies-and-installation) **|** š»[Train](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/Training.md) **|** ā[FAQ](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/FAQ.md) **|** šØ[Contribution](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/CONTRIBUTING.md)
[](https://github.com/xinntao/Real-ESRGAN/releases)
[](https://pypi.org/project/realesrgan/)
[](https://github.com/xinntao/Real-ESRGAN/issues)
[](https://github.com/xinntao/Real-ESRGAN/issues)
[](https://github.com/xinntao/Real-ESRGAN/blob/master/LICENSE)
[](https://github.com/xinntao/Real-ESRGAN/blob/master/.github/workflows/pylint.yml)
[](https://github.com/xinntao/Real-ESRGAN/blob/master/.github/workflows/publish-pip.yml)
š„ **AnimeVideo-v3 model (åØę¼«č§é¢å°ęØ”å)**. Please see [[*anime video models*](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/anime_video_model.md)] and [[*comparisons*](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/anime_comparisons.md)]
š„ **RealESRGAN_x4plus_anime_6B** for anime images **(åØę¼«ęå¾ęØ”å)**. Please see [[*anime_model*](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/anime_model.md)]
1. :boom: **Update** online Replicate demo: [](https://replicate.com/xinntao/realesrgan)
1. Online Colab demo for Real-ESRGAN: [](https://colab.research.google.com/drive/1k2Zod6kSHEvraybHl50Lys0LerhyTMCo?usp=sharing) **|** Online Colab demo for for Real-ESRGAN (**anime videos**): [](https://colab.research.google.com/drive/1yNl9ORUxxlL4N0keJa2SEPB61imPQd1B?usp=sharing)
1. Portable [Windows](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-windows.zip) / [Linux](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-ubuntu.zip) / [MacOS](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-macos.zip) **executable files for Intel/AMD/Nvidia GPU**. You can find more information [here](#portable-executable-files-ncnn). The ncnn implementation is in [Real-ESRGAN-ncnn-vulkan](https://github.com/xinntao/Real-ESRGAN-ncnn-vulkan)
Real-ESRGAN aims at developing **Practical Algorithms for General Image/Video Restoration**.
We extend the powerful ESRGAN to a practical restoration application (namely, Real-ESRGAN), which is trained with pure synthetic data.
š Thanks for your valuable feedbacks/suggestions. All the feedbacks are updated in [feedback.md](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/feedback.md).
---
If Real-ESRGAN is helpful, please help to ā this repo or recommend it to your friends š
Other recommended projects:
ā¶ļø [GFPGAN](https://github.com/TencentARC/GFPGAN): A practical algorithm for real-world face restoration
ā¶ļø [BasicSR](https://github.com/xinntao/BasicSR): An open-source image and video restoration toolbox
ā¶ļø [facexlib](https://github.com/xinntao/facexlib): A collection that provides useful face-relation functions.
ā¶ļø [HandyView](https://github.com/xinntao/HandyView): A PyQt5-based image viewer that is handy for view and comparison
ā¶ļø [HandyFigure](https://github.com/xinntao/HandyFigure): Open source of paper figures
---
### š Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data
> [[Paper](https://arxiv.org/abs/2107.10833)] [[YouTube Video](https://www.youtube.com/watch?v=fxHWoDSSvSc)] [[Bē«č®²č§£](https://www.bilibili.com/video/BV1H34y1m7sS/)] [[Poster](https://xinntao.github.io/projects/RealESRGAN_src/RealESRGAN_poster.pdf)] [[PPT slides](https://docs.google.com/presentation/d/1QtW6Iy8rm8rGLsJ0Ldti6kP-7Qyzy6XL/edit?usp=sharing&ouid=109799856763657548160&rtpof=true&sd=true)]
> [Xintao Wang](https://xinntao.github.io/), Liangbin Xie, [Chao Dong](https://scholar.google.com.hk/citations?user=OSDCB0UAAAAJ), [Ying Shan](https://scholar.google.com/citations?user=4oXBp9UAAAAJ&hl=en)
> [Tencent ARC Lab](https://arc.tencent.com/en/ai-demos/imgRestore); Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences
---
## š© Updates
- ā
Add the **realesr-general-x4v3** model - a tiny small model for general scenes. It also supports the **-dn** option to balance the noise (avoiding over-smooth results). **-dn** is short for denoising strength.
- ā
Update the **RealESRGAN AnimeVideo-v3** model. Please see [anime video models](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/anime_video_model.md) and [comparisons](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/anime_comparisons.md) for more details.
- ā
Add small models for anime videos. More details are in [anime video models](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/anime_video_model.md).
- ā
Add the ncnn implementation [Real-ESRGAN-ncnn-vulkan](https://github.com/xinntao/Real-ESRGAN-ncnn-vulkan).
- ā
Add [*RealESRGAN_x4plus_anime_6B.pth*](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth), which is optimized for **anime** images with much smaller model size. More details and comparisons with [waifu2x](https://github.com/nihui/waifu2x-ncnn-vulkan) are in [**anime_model.md**](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/anime_model.md)
- ā
Support finetuning on your own data or paired data (*i.e.*, finetuning ESRGAN). See [here](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/Training.md#Finetune-Real-ESRGAN-on-your-own-dataset)
- ā
Integrate [GFPGAN](https://github.com/TencentARC/GFPGAN) to support **face enhancement**.
- ā
Integrated to [Huggingface Spaces](https://huggingface.co/spaces) with [Gradio](https://github.com/gradio-app/gradio). See [Gradio Web Demo](https://huggingface.co/spaces/akhaliq/Real-ESRGAN). Thanks [@AK391](https://github.com/AK391)
- ā
Support arbitrary scale with `--outscale` (It actually further resizes outputs with `LANCZOS4`). Add *RealESRGAN_x2plus.pth* model.
- ā
[The inference code](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/inference_realesrgan.py) supports: 1) **tile** options; 2) images with **alpha channel**; 3) **gray** images; 4) **16-bit** images.
- ā
The training codes have been released. A detailed guide can be found in [Training.md](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/Training.md).
---
## š Demos Videos
#### Bilibili
- [大é¹å¤©å®«ēꮵ](https://www.bilibili.com/video/BV1ja41117zb)
- [Anime dance cut åØę¼«éę§čč¹](https://www.bilibili.com/video/BV1wY4y1L7hT/)
- [海蓼ēēꮵ](https://www.bilibili.com/video/BV1i3411L7Gy/)
#### YouTube
## š§ Dependencies and Installation
- Python >= 3.7 (Recommend to use [Anaconda](https://www.anaconda.com/download/#linux) or [Miniconda](https://docs.conda.io/en/latest/miniconda.html))
- [PyTorch >= 1.7](https://pytorch.org/)
### Installation
1. Clone repo
```bash
git clone https://github.com/xinntao/Real-ESRGAN.git
cd Real-ESRGAN
```
1. Install dependent packages
```bash
# Install basicsr - https://github.com/xinntao/BasicSR
# We use BasicSR for both training and inference
pip install basicsr
# facexlib and gfpgan are for face enhancement
pip install facexlib
pip install gfpgan
pip install -r requirements.txt
python setup.py develop
```
---
## ā” Quick Inference
There are usually three ways to inference Real-ESRGAN.
1. [Online inference](#online-inference)
1. [Portable executable files (NCNN)](#portable-executable-files-ncnn)
1. [Python script](#python-script)
### Online inference
1. You can try in our website: [ARC Demo](https://arc.tencent.com/en/ai-demos/imgRestore) (now only support RealESRGAN_x4plus_anime_6B)
1. [Colab Demo](https://colab.research.google.com/drive/1k2Zod6kSHEvraybHl50Lys0LerhyTMCo?usp=sharing) for Real-ESRGAN **|** [Colab Demo](https://colab.research.google.com/drive/1yNl9ORUxxlL4N0keJa2SEPB61imPQd1B?usp=sharing) for Real-ESRGAN (**anime videos**).
### Portable executable files (NCNN)
You can download [Windows](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-windows.zip) / [Linux](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-ubuntu.zip) / [MacOS](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-macos.zip) **executable files for Intel/AMD/Nvidia GPU**.
This executable file is **portable** and includes all the binaries and models required. No CUDA or PyTorch environment is needed.
You can simply run the following command (the Windows example, more information is in the README.md of each executable files):
```bash
./realesrgan-ncnn-vulkan.exe -i input.jpg -o output.png -n model_name
```
We have provided five models:
1. realesrgan-x4plus (default)
2. realesrnet-x4plus
3. realesrgan-x4plus-anime (optimized for anime images, small model size)
4. realesr-animevideov3 (animation video)
You can use the `-n` argument for other models, for example, `./realesrgan-ncnn-vulkan.exe -i input.jpg -o output.png -n realesrnet-x4plus`
#### Usage of portable executable files
1. Please refer to [Real-ESRGAN-ncnn-vulkan](https://github.com/xinntao/Real-ESRGAN-ncnn-vulkan#computer-usages) for more details.
1. Note that it does not support all the functions (such as `outscale`) as the python script `inference_realesrgan.py`.
```console
Usage: realesrgan-ncnn-vulkan.exe -i infile -o outfile [options]...
-h show this help
-i input-path input image path (jpg/png/webp) or directory
-o output-path output image path (jpg/png/webp) or directory
-s scale upscale ratio (can be 2, 3, 4. default=4)
-t tile-size tile size (>=32/0=auto, default=0) can be 0,0,0 for multi-gpu
-m model-path folder path to the pre-trained models. default=models
-n model-name model name (default=realesr-animevideov3, can be realesr-animevideov3 | realesrgan-x4plus | realesrgan-x4plus-anime | realesrnet-x4plus)
-g gpu-id gpu device to use (default=auto) can be 0,1,2 for multi-gpu
-j load:proc:save thread count for load/proc/save (default=1:2:2) can be 1:2,2,2:2 for multi-gpu
-x enable tta mode"
-f format output image format (jpg/png/webp, default=ext/png)
-v verbose output
```
Note that it may introduce block inconsistency (and also generate slightly different results from the PyTorch implementation), because this executable file first crops the input image into several tiles, and then processes them separately, finally stitches together.
### Python script
#### Usage of python script
1. You can use X4 model for **arbitrary output size** with the argument `outscale`. The program will further perform cheap resize operation after the Real-ESRGAN output.
```console
Usage: python inference_realesrgan.py -n RealESRGAN_x4plus -i infile -o outfile [options]...
A common command: python inference_realesrgan.py -n RealESRGAN_x4plus -i infile --outscale 3.5 --face_enhance
-h show this help
-i --input Input image or folder. Default: inputs
-o --output Output folder. Default: results
-n --model_name Model name. Default: RealESRGAN_x4plus
-s, --outscale The final upsampling scale of the image. Default: 4
--suffix Suffix of the restored image. Default: out
-t, --tile Tile size, 0 for no tile during testing. Default: 0
--face_enhance Whether to use GFPGAN to enhance face. Default: False
--fp32 Use fp32 precision during inference. Default: fp16 (half precision).
--ext Image extension. Options: auto | jpg | png, auto means using the same extension as inputs. Default: auto
```
#### Inference general images
Download pre-trained models: [RealESRGAN_x4plus.pth](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth)
```bash
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth -P weights
```
Inference!
```bash
python inference_realesrgan.py -n RealESRGAN_x4plus -i inputs --face_enhance
```
Results are in the `results` folder
#### Inference anime images
Pre-trained models: [RealESRGAN_x4plus_anime_6B](https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth)
More details and comparisons with [waifu2x](https://github.com/nihui/waifu2x-ncnn-vulkan) are in [**anime_model.md**](https://raw.githubusercontent.com/xinntao/Real-ESRGAN/a4abfb2979a7bbff3f69f58f58ae324608821e27/docs/anime_model.md)
```bash
# download model
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth -P weights
# inference
python inference_realesrgan.py -n RealESRGAN_x4plus_anime_6B -i inputs
```
Results are in the `results` folder
---
## BibTeX
@InProceedings{wang2021realesrgan,
author = {Xintao Wang and Liangbin Xie and Chao Dong and Ying Shan},
title = {Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data},
booktitle = {International Conference on Computer Vision Workshops (ICCVW)},
date = {2021}
}
## š§ Contact
If you have any question, please email `xintao.wang@outlook.com` or `xintaowang@tencent.com`.
## š§© Projects that use Real-ESRGAN
If you develop/use Real-ESRGAN in your projects, welcome to let me know.
- NCNN-Android: [RealSR-NCNN-Android](https://github.com/tumuyan/RealSR-NCNN-Android) by [tumuyan](https://github.com/tumuyan)
- VapourSynth: [vs-realesrgan](https://github.com/HolyWu/vs-realesrgan) by [HolyWu](https://github.com/HolyWu)
- NCNN: [Real-ESRGAN-ncnn-vulkan](https://github.com/xinntao/Real-ESRGAN-ncnn-vulkan)
**GUI**
- [Waifu2x-Extension-GUI](https://github.com/AaronFeng753/Waifu2x-Extension-GUI) by [AaronFeng753](https://github.com/AaronFeng753)
- [Squirrel-RIFE](https://github.com/Justin62628/Squirrel-RIFE) by [Justin62628](https://github.com/Justin62628)
- [Real-GUI](https://github.com/scifx/Real-GUI) by [scifx](https://github.com/scifx)
- [Real-ESRGAN_GUI](https://github.com/net2cn/Real-ESRGAN_GUI) by [net2cn](https://github.com/net2cn)
- [Real-ESRGAN-EGUI](https://github.com/WGzeyu/Real-ESRGAN-EGUI) by [WGzeyu](https://github.com/WGzeyu)
- [anime_upscaler](https://github.com/shangar21/anime_upscaler) by [shangar21](https://github.com/shangar21)
- [Upscayl](https://github.com/upscayl/upscayl) by [Nayam Amarshe](https://github.com/NayamAmarshe) and [TGS963](https://github.com/TGS963)
## š¤ Acknowledgement
Thanks for all the contributors.
- [AK391](https://github.com/AK391): Integrate RealESRGAN to [Huggingface Spaces](https://huggingface.co/spaces) with [Gradio](https://github.com/gradio-app/gradio). See [Gradio Web Demo](https://huggingface.co/spaces/akhaliq/Real-ESRGAN).
- [Asiimoviet](https://github.com/Asiimoviet): Translate the README.md to Chinese (äøę).
- [2ji3150](https://github.com/2ji3150): Thanks for the [detailed and valuable feedbacks/suggestions](https://github.com/xinntao/Real-ESRGAN/issues/131).
- [Jared-02](https://github.com/Jared-02): Translate the Training.md to Chinese (äøę).