Real-ESRGAN aims at developing Practical Algorithms for General Image Restoration.

Xintao

Last update: Jan 2, 2023

Related tags

Overview

Real-ESRGAN

Colab Demo for Real-ESRGAN .
Portable Windows executable file. You can find more information here.

Real-ESRGAN aims at developing Practical Algorithms for General Image Restoration.
We extend the powerful ESRGAN to a practical restoration application (namely, Real-ESRGAN), which is trained with pure synthetic data.

🚩 The training codes have been released. A detailed guide can be found in Training.md.

📖 Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data

[Paper] [Project Page] [Demo]
Xintao Wang, Liangbin Xie, Chao Dong, Ying Shan
Applied Research Center (ARC), Tencent PCG
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences

We have provided a pretrained model (RealESRGAN_x4plus.pth) with upsampling X4.
Note that RealESRGAN may still fail in some cases as the real-world degradations are really too complex.
Moreover, it may not perform well on human faces, text, etc, which will be optimized later.

Real-ESRGAN will be a long-term supported project (in my current plan 😃 ). It will be continuously updated in my spare time.

Here is a TODO list in the near future:

optimize for human faces
optimize for texts
optimize for animation images
support more scales
support controllable restoration strength

If you have any good ideas or demands, please open an issue/discussion to let me know.
If you have some images that Real-ESRGAN could not well restored, please also open an issue/discussion. I will record it (but I cannot guarantee to resolve it 😛 ). If necessary, I will open a page to specially record these real-world cases that need to be solved, but the current technology is difficult to handle well.

Portable executable files

You can download Windows executable files from https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.1/RealESRGAN-ncnn-vulkan-20210725-windows.zip

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:

./realesrgan-ncnn-vulkan.exe -i input.jpg -o output.png

We have provided three models:

realesrgan-x4plus (default)
realesrnet-x4plus
esrgan-x4

You can use the -n argument for other models, for example, ./realesrgan-ncnn-vulkan.exe -i input.jpg -o output.png -n realesrnet-x4plus

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.

This executable file is based on the wonderful Tencent/ncnn and realsr-ncnn-vulkan by nihui.

🔧 Dependencies and Installation

Python >= 3.7 (Recommend to use Anaconda or Miniconda)
PyTorch >= 1.7

Installation

Clone repo

git clone https://github.com/xinntao/Real-ESRGAN.git
cd Real-ESRGAN

Install dependent packages

# Install basicsr - https://github.com/xinntao/BasicSR
# We use BasicSR for both training and inference
pip install basicsr
pip install -r requirements.txt

⚡ Quick Inference

Download pre-trained models: RealESRGAN_x4plus.pth

Download pretrained models:

wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth -P experiments/pretrained_models

Inference!

python inference_realesrgan.py --model_path experiments/pretrained_models/RealESRGAN_x4plus.pth --input inputs

Results are in the results folder

🏰 Model Zoo

💻 Training

A detailed guide can be found in Training.md.

BibTeX

@Article{wang2021realesrgan,
    title={Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data},
    author={Xintao Wang and Liangbin Xie and Chao Dong and Ying Shan},
    journal={arXiv:2107.10833},
    year={2021}
}

📧 Contact

If you have any question, please email [email protected] or [email protected].

Comments

Unable to use my own trained models.

Hello,

I updated to the last repo version and now I'm not able to use the models I have just trained.

Before I used the argument --input_path to indicate the trained model path. Now --input_path exists no more and I'm moving the trained models to rsgan/weights as indicates the new inference_realsrgan.py but then I get an error like the following:

Traceback (most recent call last): File "inference_realesrgan.py", line 128, in main() File "inference_realesrgan.py", line 75, in main scale=netscale, UnboundLocalError: local variable 'netscale' referenced before assignment

Any idea how should I use the weights obtained by training or fine-tuning with my own data?

opened by SirSykon 11
refactor inference_realesrgan_video.py
Avoid outputs temporary file by streaming processing. remove some feature:

image/folder input

image/folder output

timer. tqdm can do it

input argument change:

add 'ffmpeg_bin'

remove 'video'

remove 'audio'

remove 'ext'

issue #212
opened by Juszoe 9

关于输出的图片全为黑色的问题

一、运行环境： 操作系统：Windows 11 专业版 21H2 处理器：Intel(R) Core(TM) i5-9400F CPU @ 2.90GHz 2.90 GHz 显卡信息：NVIDIA GeForce RTX 1660Ti

(gfpgan) > nvidia-smi
Sat May 14 16:53:18 2022
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 512.59       Driver Version: 512.59       CUDA Version: 11.6     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ... WDDM  | 00000000:01:00.0 Off |                  N/A |
| 34%   30C    P8    18W / 130W |    509MiB /  6144MiB |      2%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A       576    C+G   ...n1h2txyewy\SearchHost.exe    N/A      |
|    0   N/A  N/A      1552    C+G   C:\Windows\System32\dwm.exe     N/A      |
|    0   N/A  N/A      3100    C+G   C:\Windows\explorer.exe         N/A      |
|    0   N/A  N/A      3376    C+G   ...artMenuExperienceHost.exe    N/A      |
|    0   N/A  N/A      3896    C+G   ...8bbwe\WindowsTerminal.exe    N/A      |
|    0   N/A  N/A      4120    C+G   C:\Windows\System32\dwm.exe     N/A      |
|    0   N/A  N/A      5608    C+G   ...y\ShellExperienceHost.exe    N/A      |
|    0   N/A  N/A      6516    C+G   ...dows\System32\LogonUI.exe    N/A      |
|    0   N/A  N/A      6636    C+G   ...ge\Application\msedge.exe    N/A      |
|    0   N/A  N/A      7324    C+G   ...2txyewy\TextInputHost.exe    N/A      |
|    0   N/A  N/A      8444    C+G   ...210.39\msedgewebview2.exe    N/A      |
|    0   N/A  N/A     10088    C+G   ...ows\System32\WUDFHost.exe    N/A      |
|    0   N/A  N/A     10160    C+G   ...lPanel\SystemSettings.exe    N/A      |
|    0   N/A  N/A     10424    C+G   ...ekyb3d8bbwe\YourPhone.exe    N/A      |
|    0   N/A  N/A     12744    C+G   ...210.39\msedgewebview2.exe    N/A      |
+-----------------------------------------------------------------------------+

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Tue_Mar__8_18:36:24_Pacific_Standard_Time_2022
Cuda compilation tools, release 11.6, V11.6.124
Build cuda_11.6.r11.6/compiler.31057947_0

二、相关软件 Anaconda 4.10.3
Python 3.8.13 pytorch 1.11.0

三、使用过程 最开始是使用 GFPGAN 项目，按教程安装好，用示例的命令（python inference_gfpgan.py -i inputs/whole_imgs -o results -v 1.3 -s 2）进行测试，脸部分割、优化正常，但是输出的 restored_imgs 只有脸部，背景全为黑色。

关闭-bg_upsampler(指定一个不存在的smapler)后，restored_imgs里的图片输出正常，脸部得到优化，但背景还是模糊的（符合预期，因为没有使用背景优化），确定为realesrgan运行不正常。

下载了Real-ESRGAN的绿色版，功能正常，不会出现图片为黑色的情况。

克隆此项目的代码，按教程安装，用示例的命令（python inference_realesrgan.py -n RealESRGAN_x4plus -i inputs --face_enhance）进行测试，输出的图片全为黑色。

为排除操作系统问题，在wsl2下安装Ubuntu 20.04.4 LTS，再次进行上面的测试，问题一样。

搜索了issues，没有发现相关的问题，希望能得到帮助。

非常感谢，这个几个项目太棒了！

opened by rayeesoft 8

盲道中 sinc模块做振铃和过冲伪像

因为我像看看sinc 模块最终生成图像的样子,所以我想将sinc 模块提取如下我想将sinc 模块提出来, 删除了作者的注释,现在如下,可是我cv2.imshow的结果图却是白板,您能帮我瞅瞅我哪儿的处理做错了

def circular_lowpass_kernel(cutoff, kernel_size, pad_to=0):
    assert kernel_size % 2 == 1, 'Kernel size must be an odd number.'
    kernel = np.fromfunction(
        lambda x, y: cutoff * special.j1(cutoff * np.sqrt(
            (x - (kernel_size - 1) / 2) ** 2 + (y - (kernel_size - 1) / 2) ** 2)) / (2 * np.pi * np.sqrt(
            (x - (kernel_size - 1) / 2) ** 2 + (y - (kernel_size - 1) / 2) ** 2)), [kernel_size, kernel_size])
    kernel[(kernel_size - 1) // 2, (kernel_size - 1) // 2] = cutoff ** 2 / (4 * np.pi)
    kernel = kernel / np.sum(kernel)
    if pad_to > kernel_size:
        pad_size = (pad_to - kernel_size) // 2
        kernel = np.pad(kernel, ((pad_size, pad_size), (pad_size, pad_size)))
    return kernel


def filter2D(img, kernel):
    img = torch.FloatTensor(img)
    kernel = torch.FloatTensor(kernel)
    k = kernel.shape[-1]
    assert k % 2 == 1, 'Kernel size must be an odd number.'
    h, w, c = img.shape
    img = img.view(1, c, h, w)
    if k % 2 == 1:
        img = F.pad(img, (k // 2, k // 2, k // 2, k // 2), mode='reflect')  # padding
    else:
        raise ValueError('Wrong kernel size')
    ph, pw = img.size()[-2:]
    device = torch.device('cuda')
    if kernel.size(0) == 1:
        # apply the same kenrel to all batch images
        img = img.view(1 * c, 1, ph, pw).to(device)
        kernel = kernel.view(1, 1, k, k).to(device)
        return F.conv2d(img, kernel, padding=0).view(h, w, c).cpu().numpy().clip(0, 255)
    else:
        img = img.view(1, c, ph, pw).to(device)
        kernel = kernel.view(1, 1, k, k).repeat(1, c, 1, 1).view(c, 1, k, k).to(device)
        return F.conv2d(img, kernel, groups=c).view(h, w, c).cpu().numpy().clip(0, 255)
img = cv2.imread("./1111.jpg")
sinc_kernel_size = random.choice([7, 9, 11, 13, 15, 17, 19, 21])
omega_c = np.random.uniform(np.pi / 3, np.pi)
sinc_kernel = circular_lowpass_kernel(omega_c, sinc_kernel_size, pad_to=21)
cv2.imshow("sinc_kernel", sinc_kernel)
cv2.imshow("img0", img)
res_img = filter2D(img, sinc_kernel)
cv2.imshow("res_img", res_img)
cv2.waitKey(0)

opened by tongchangD 8

Error related to File/Video Import Widget

MessageError Traceback (most recent call last) in () 14 15 # upload images ---> 16 uploaded = files.upload() 17 for filename in uploaded.keys(): 18 dst_path = os.path.join(upload_folder, filename)

2 frames /usr/local/lib/python3.7/dist-packages/google/colab/_message.py in read_reply_from_input(message_id, timeout_sec) 104 reply.get('colab_msg_id') == message_id): 105 if 'error' in reply: --> 106 raise MessageError(reply['error']) 107 return reply.get('data', None) 108

MessageError: TypeError: Cannot read property '_uploadFiles' of undefined

opened by Ghee36 8
cannot import name 'AvgTimer' from 'basicsr.utils.logger

python inference_realesrgan_video.py -i inputs/video/onepiece_demo.mp4 -n RealESRGANv2-animevideo-xsx2 -s 2 - v -a --half --suffix outx2 Traceback (most recent call last): File "D:\python学习\Real-ESRGAN-master\Real-ESRGAN-master\inference_realesrgan_video.py", line 9, in from basicsr.utils.logger import AvgTimer ImportError: cannot import name 'AvgTimer' from 'basicsr.utils.logger' (D:\Python\lib\site-packages\basicsr\utils\logger.py)

opened by wsysl1989 7
The usage of _dequeue_and_enqueue function in RealESRNetModel

Hi, I read the code serveral times, but cannot figure out what the role of _dequeue_and_enqueue function in RealESRNetModel is. This function is only used in feed_data(), which just put self.lq and self.gt into self.queue_lq and self.queue_gt. But I cannot find some other codes to use self.queue_lq and self.queue_gt. I would appreciate it if someone could explain this.

opened by YangGangZhiQi 7
'VCOMP140D.DLL' is required

A straightforward execution of the 'Windows executable files' fails because 'VCOMP140D.DLL' is required.

For me it was necessary to install 'Visual Studio 2019' and load the 'MSVC v142' package to solve this problem. Putting the 'VCOMP140D.DLL' into the 'Windows executable files' would help other users.

The result look good, thanks for releasing.

opened by ghost 7
ESRGAN massive CPU bottleneck

Like I mentioned in this discussion: https://github.com/xinntao/Real-ESRGAN/discussions/269 ESRGAN is severely bottlenecked by my beastly CPU even though my GPU is trash. Ryzen 9 5900x: GTX 1080: My GPU is load is currently at 19% average while CPU is around 90-95%. It's literally telling me it's IDLING: This is using the portable NCNN version with realesr-animevideov3 model at 4x, going from 628x480 to 2512x1920.

opened by TechnoMasterBoy 6

Portable Executable error: Invalid bitcast %379 = bitcast i32 %378 to i16

I've installed on my linux host portable executable, tried to run it and received that error:

quantum0@test-host:~/realesrgan$ ./realesrgan-ncnn-vulkan -i input.jpg -o output.png -n realesrgan-x4plus
[0 llvmpipe (LLVM 12.0.0, 256 bits)]  queueC=0[1]  queueG=0[1]  queueT=0[1]
[0 llvmpipe (LLVM 12.0.0, 256 bits)]  bugsbn1=0  bugbilz=0  bugcopc=0  bugihfa=0
[0 llvmpipe (LLVM 12.0.0, 256 bits)]  fp16-p/s/a=1/1/0  int8-p/s/a=1/1/0
[0 llvmpipe (LLVM 12.0.0, 256 bits)]  subgroup=8  basic=1  vote=1  ballot=1  shuffle=0
WARNING: lavapipe is not a conformant vulkan implementation, testing use only.
Invalid bitcast
  %379 = bitcast i32 %378 to i16
LLVM ERROR: Broken function
Aborted

What I'm doing wrong?

Maybe this will be useful:

quantum0@test-host:~/realesrgan$ uname -a
Linux test-host 5.4.0-96-generic #109-Ubuntu SMP Wed Jan 12 16:49:16 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux

Btw it's not enough just to download executable and run, i had to install these dependencies:

sudo apt-get install libvulkan-dev
sudo apt-get install libgomp1

opened by Quantum-0 6

图片图块排版不正确，请修复

我使用windows的版本，程序目录下自带的input.jpg测试，使用的指令如下 D:\Real-ESRGAN-realesrgan-ncnn-vulkan-20211212-windows\realesrgan-ncnn-vulkan.exe -i input.jpg -o test.png -s 2 D:\Real-ESRGAN-realesrgan-ncnn-vulkan-20211212-windows\realesrgan-ncnn-vulkan.exe -i input.jpg -o test.png -s 1 分别测试了1倍数和2倍数，都是这样的排版，请修复

opened by cuihua0 6
realesr-general-wdn-x4v3如何训练的

你好，我看github上更新了模型realesr-general-x4v3，通过 -dn参数可以调节噪声抑制的水平。需要2个模型realesr-general-wdn-x4v3.pth和realesr-general-x4v3.pth。其中realesr-general-wdn-x4v3.pth应该是没有噪声抑制的模型，请问这个模型是怎么训练出来的呢？谢谢

opened by buwangchuxin1992 0
Output file when running -s argument shows corruption of data

I am running: Real-ESRGAN ncnn Vulkan and the -s argument only works for the default realesr-animevideov3 model and not for anything other that is provided

all other models are stuck at 4x scale

Steps to recreate:

realesrgan-ncnn-vulkan.exe -i input.jpg -o output.png -n realesrgan-x4plus -s 2

or

realesrgan-ncnn-vulkan.exe -i input.jpg -o output.png -n realesrgan-x4plus-anime -s 2

Output (for both):

opened by AlphaHasher 0

invalid bitcast error

I get a very similar output to #262, even after installing the packages recommended in the comments on that issue:

$ ./realesrgan-ncnn-vulkan -i input.jpg -o output.png -n realesr-animevideov3 -s 2
[0 llvmpipe (LLVM 12.0.0, 256 bits)]  queueC=0[1]  queueG=0[1]  queueT=0[1]
[0 llvmpipe (LLVM 12.0.0, 256 bits)]  bugsbn1=0  bugbilz=0  bugcopc=0  bugihfa=0
[0 llvmpipe (LLVM 12.0.0, 256 bits)]  fp16-p/s/a=1/1/0  int8-p/s/a=1/1/0
[0 llvmpipe (LLVM 12.0.0, 256 bits)]  subgroup=8  basic=1  vote=1  ballot=1  shuffle=0
WARNING: lavapipe is not a conformant vulkan implementation, testing use only.
Invalid bitcast
  %379 = bitcast i32 %378 to i16
LLVM ERROR: Broken function
Aborted (core dumped)

nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 515.86.01    Driver Version: 515.86.01    CUDA Version: 11.7     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:02:00.0 Off |                  N/A |
| 33%   27C    P8     9W / 190W |      0MiB /  8192MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

vulkaninfo

ERROR: [Loader Message] Code 0 : /usr/lib/i386-linux-gnu/libvulkan_lvp.so: wrong ELF class: ELFCLASS32
ERROR: [Loader Message] Code 0 : /usr/lib/i386-linux-gnu/libvulkan_radeon.so: wrong ELF class: ELFCLASS32
ERROR: [Loader Message] Code 0 : /usr/lib/i386-linux-gnu/libvulkan_intel.so: wrong ELF class: ELFCLASS32
'DISPLAY' environment variable not set... skipping surface info
WARNING: lavapipe is not a conformant vulkan implementation, testing use only.
==========
VULKANINFO
==========

Vulkan Instance Version: 1.2.131


Instance Extensions: count = 18
====================
	VK_EXT_acquire_xlib_display            : extension revision 1
	VK_EXT_debug_report                    : extension revision 10
	VK_EXT_debug_utils                     : extension revision 1
	VK_EXT_direct_mode_display             : extension revision 1
	VK_EXT_display_surface_counter         : extension revision 1
	VK_KHR_device_group_creation           : extension revision 1
	VK_KHR_display                         : extension revision 23
	VK_KHR_external_fence_capabilities     : extension revision 1
	VK_KHR_external_memory_capabilities    : extension revision 1
	VK_KHR_external_semaphore_capabilities : extension revision 1
	VK_KHR_get_display_properties2         : extension revision 1
	VK_KHR_get_physical_device_properties2 : extension revision 2
	VK_KHR_get_surface_capabilities2       : extension revision 1
	VK_KHR_surface                         : extension revision 25
	VK_KHR_surface_protected_capabilities  : extension revision 1
	VK_KHR_wayland_surface                 : extension revision 6
	VK_KHR_xcb_surface                     : extension revision 6
	VK_KHR_xlib_surface                    : extension revision 6

Layers: count = 3
=======
VK_LAYER_LUNARG_standard_validation (LunarG Standard Validation Layer) Vulkan version 1.0.131, layer version 1:
	Layer Extensions: count = 0
	Devices: count = 1
		GPU id 	: 0 (llvmpipe (LLVM 12.0.0, 256 bits))
		Layer-Device Extensions: count = 0

VK_LAYER_MESA_device_select (Linux device selection layer) Vulkan version 1.2.73, layer version 1:
	Layer Extensions: count = 0
	Devices: count = 1
		GPU id 	: 0 (llvmpipe (LLVM 12.0.0, 256 bits))
		Layer-Device Extensions: count = 0

VK_LAYER_MESA_overlay (Mesa Overlay layer) Vulkan version 1.1.73, layer version 1:
	Layer Extensions: count = 0
	Devices: count = 1
		GPU id 	: 0 (llvmpipe (LLVM 12.0.0, 256 bits))
		Layer-Device Extensions: count = 0

Presentable Surfaces:
=====================

Groups:
=======
	Device Group Properties (Group 0):
		physicalDeviceCount: count = 1
			llvmpipe (LLVM 12.0.0, 256 bits) (ID: 0)
		subsetAllocation = 0

	Device Group Present Capabilities (Group 0):
WARNING: lavapipe is not a conformant vulkan implementation, testing use only.
		llvmpipe (LLVM 12.0.0, 256 bits) (ID: 0)
		Can present images from the following devices:
			llvmpipe (LLVM 12.0.0, 256 bits) (ID: 0)
		Present modes:
			DEVICE_GROUP_PRESENT_MODE_LOCAL_BIT_KHR


Device Properties and Extensions:
=================================
GPU0:
VkPhysicalDeviceProperties:
---------------------------
	apiVersion     = 4198582 (1.1.182)
	driverVersion  = 1 (0x0001)
	vendorID       = 0x10005
	deviceID       = 0x0000
	deviceType     = PHYSICAL_DEVICE_TYPE_CPU
	deviceName     = llvmpipe (LLVM 12.0.0, 256 bits)

VkPhysicalDeviceLimits:
-----------------------
	maxImageDimension1D                             = 16384
	maxImageDimension2D                             = 16384
	maxImageDimension3D                             = 4096
	maxImageDimensionCube                           = 32768
	maxImageArrayLayers                             = 2048
	maxTexelBufferElements                          = 134217728
	maxUniformBufferRange                           = 65536
	maxStorageBufferRange                           = 134217728
	maxPushConstantsSize                            = 128
	maxMemoryAllocationCount                        = 4294967295
	maxSamplerAllocationCount                       = 32768
	bufferImageGranularity                          = 0x00000040
	sparseAddressSpaceSize                          = 0x00000000
	maxBoundDescriptorSets                          = 8
	maxPerStageDescriptorSamplers                   = 32
	maxPerStageDescriptorUniformBuffers             = 15
	maxPerStageDescriptorStorageBuffers             = 16
	maxPerStageDescriptorSampledImages              = 128
	maxPerStageDescriptorStorageImages              = 16
	maxPerStageDescriptorInputAttachments           = 8
	maxPerStageResources                            = 128
	maxDescriptorSetSamplers                        = 32768
	maxDescriptorSetUniformBuffers                  = 256
	maxDescriptorSetUniformBuffersDynamic           = 256
	maxDescriptorSetStorageBuffers                  = 256
	maxDescriptorSetStorageBuffersDynamic           = 256
	maxDescriptorSetSampledImages                   = 256
	maxDescriptorSetStorageImages                   = 256
	maxDescriptorSetInputAttachments                = 256
	maxVertexInputAttributes                        = 32
	maxVertexInputBindings                          = 32
	maxVertexInputAttributeOffset                   = 2047
	maxVertexInputBindingStride                     = 2048
	maxVertexOutputComponents                       = 128
	maxTessellationGenerationLevel                  = 64
	maxTessellationPatchSize                        = 32
	maxTessellationControlPerVertexInputComponents  = 128
	maxTessellationControlPerVertexOutputComponents = 128
	maxTessellationControlPerPatchOutputComponents  = 128
	maxTessellationControlTotalOutputComponents     = 4096
	maxTessellationEvaluationInputComponents        = 128
	maxTessellationEvaluationOutputComponents       = 128
	maxGeometryShaderInvocations                    = 32
	maxGeometryInputComponents                      = 64
	maxGeometryOutputComponents                     = 128
	maxGeometryOutputVertices                       = 1024
	maxGeometryTotalOutputComponents                = 1024
	maxFragmentInputComponents                      = 128
	maxFragmentOutputAttachments                    = 8
	maxFragmentDualSrcAttachments                   = 2
	maxFragmentCombinedOutputResources              = 8
	maxComputeSharedMemorySize                      = 32768
	maxComputeWorkGroupCount: count = 3
		65535
		65535
		65535
	maxComputeWorkGroupInvocations                  = 1024
	maxComputeWorkGroupSize: count = 3
		1024
		1024
		1024
	subPixelPrecisionBits                           = 8
	subTexelPrecisionBits                           = 8
	mipmapPrecisionBits                             = 8
	maxDrawIndexedIndexValue                        = 4294967295
	maxDrawIndirectCount                            = 4294967295
	maxSamplerLodBias                               = 16
	maxSamplerAnisotropy                            = 16
	maxViewports                                    = 16
	maxViewportDimensions: count = 2
		16384
		16384
	viewportBoundsRange: count = 2
		-32768
		32768
	viewportSubPixelBits                            = 0
	minMemoryMapAlignment                           = 64
	minTexelBufferOffsetAlignment                   = 0x00000010
	minUniformBufferOffsetAlignment                 = 0x00000010
	minStorageBufferOffsetAlignment                 = 0x00000010
	minTexelOffset                                  = -32
	maxTexelOffset                                  = 31
	minTexelGatherOffset                            = -32
	maxTexelGatherOffset                            = 31
	minInterpolationOffset                          = -2
	maxInterpolationOffset                          = 2
	subPixelInterpolationOffsetBits                 = 8
	maxFramebufferWidth                             = 16384
	maxFramebufferHeight                            = 16384
	maxFramebufferLayers                            = 2048
	framebufferColorSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	framebufferDepthSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	framebufferStencilSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	framebufferNoAttachmentsSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	maxColorAttachments                             = 8
	sampledImageColorSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	sampledImageIntegerSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	sampledImageDepthSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	sampledImageStencilSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	storageImageSampleCounts:
		SAMPLE_COUNT_1_BIT
		SAMPLE_COUNT_4_BIT
	maxSampleMaskWords                              = 1
	timestampComputeAndGraphics                     = true
	timestampPeriod                                 = 1
	maxClipDistances                                = 8
	maxCullDistances                                = 8
	maxCombinedClipAndCullDistances                 = 8
	discreteQueuePriorities                         = 2
	pointSizeRange: count = 2
		0
		255
	lineWidthRange: count = 2
		1
		255
	pointSizeGranularity                            = 0.125
	lineWidthGranularity                            = 0.0078125
	strictLines                                     = true
	standardSampleLocations                         = true
	optimalBufferCopyOffsetAlignment                = 0x00000080
	optimalBufferCopyRowPitchAlignment              = 0x00000080
	nonCoherentAtomSize                             = 0x00000040

VkPhysicalDeviceSparseProperties:
---------------------------------
	residencyStandard2DBlockShape            = false
	residencyStandard2DMultisampleBlockShape = false
	residencyStandard3DBlockShape            = false
	residencyAlignedMipSize                  = false
	residencyNonResidentStrict               = false

VkPhysicalDeviceDriverPropertiesKHR:
------------------------------------
	driverID           = UNKNOWN_VkDriverId
	driverName         = llvmpipe
	driverInfo         = Mesa 21.2.6 (LLVM 12.0.0)
	conformanceVersion = 1.0.0.0

VkPhysicalDeviceIDProperties:
-----------------------------
	deviceUUID      = 00000000-0000-0000-0000-000000000000
	driverUUID      = 00000000-0000-0000-0000-000000000000
	deviceNodeMask  = 0
	deviceLUIDValid = false

VkPhysicalDeviceLineRasterizationPropertiesEXT:
-----------------------------------------------
	lineSubPixelPrecisionBits = 8

VkPhysicalDeviceMaintenance3Properties:
---------------------------------------
	maxPerSetDescriptors    = 1024
	maxMemoryAllocationSize = 0x80000000

VkPhysicalDeviceMultiviewProperties:
------------------------------------
	maxMultiviewViewCount     = 6
	maxMultiviewInstanceIndex = 2147483647

VkPhysicalDevicePointClippingProperties:
----------------------------------------
	pointClippingBehavior = POINT_CLIPPING_BEHAVIOR_ALL_CLIP_PLANES

VkPhysicalDeviceProtectedMemoryProperties:
------------------------------------------
	protectedNoFault = false

VkPhysicalDevicePushDescriptorPropertiesKHR:
--------------------------------------------
	maxPushDescriptors = 32

VkPhysicalDeviceSamplerFilterMinmaxPropertiesEXT:
-------------------------------------------------
	filterMinmaxSingleComponentFormats = true
	filterMinmaxImageComponentMapping  = true

VkPhysicalDeviceSubgroupProperties:
-----------------------------------
	subgroupSize              = 8
	supportedStages:
		SHADER_STAGE_FRAGMENT_BIT
		SHADER_STAGE_COMPUTE_BIT
		SHADER_STAGE_ALL_GRAPHICS
		SHADER_STAGE_ALL
	supportedOperations:
		SUBGROUP_FEATURE_BASIC_BIT
		SUBGROUP_FEATURE_VOTE_BIT
		SUBGROUP_FEATURE_ARITHMETIC_BIT
		SUBGROUP_FEATURE_BALLOT_BIT
	quadOperationsInAllStages = false

VkPhysicalDeviceTransformFeedbackPropertiesEXT:
-----------------------------------------------
	maxTransformFeedbackStreams                = 4
	maxTransformFeedbackBuffers                = 4
	maxTransformFeedbackBufferSize             = 0xffffffff
	maxTransformFeedbackStreamDataSize         = 512
	maxTransformFeedbackBufferDataSize         = 512
	maxTransformFeedbackBufferDataStride       = 512
	transformFeedbackQueries                   = true
	transformFeedbackStreamsLinesTriangles     = false
	transformFeedbackRasterizationStreamSelect = false
	transformFeedbackDraw                      = true

VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT:
----------------------------------------------------
	maxVertexAttribDivisor = 4294967295


Device Extensions: count = 54
------------------
	VK_EXT_calibrated_timestamps          : extension revision 2
	VK_EXT_conditional_rendering          : extension revision 2
	VK_EXT_custom_border_color            : extension revision 12
	VK_EXT_extended_dynamic_state         : extension revision 1
	VK_EXT_extended_dynamic_state2        : extension revision 1
	VK_EXT_host_query_reset               : extension revision 1
	VK_EXT_index_type_uint8               : extension revision 1
	VK_EXT_line_rasterization             : extension revision 1
	VK_EXT_multi_draw                     : extension revision 1
	VK_EXT_post_depth_coverage            : extension revision 1
	VK_EXT_private_data                   : extension revision 1
	VK_EXT_provoking_vertex               : extension revision 1
	VK_EXT_sampler_filter_minmax          : extension revision 2
	VK_EXT_scalar_block_layout            : extension revision 1
	VK_EXT_separate_stencil_usage         : extension revision 1
	VK_EXT_shader_stencil_export          : extension revision 1
	VK_EXT_shader_viewport_index_layer    : extension revision 1
	VK_EXT_transform_feedback             : extension revision 1
	VK_EXT_vertex_attribute_divisor       : extension revision 3
	VK_EXT_vertex_input_dynamic_state     : extension revision 2
	VK_GOOGLE_decorate_string             : extension revision 1
	VK_GOOGLE_hlsl_functionality1         : extension revision 1
	VK_KHR_16bit_storage                  : extension revision 1
	VK_KHR_8bit_storage                   : extension revision 1
	VK_KHR_bind_memory2                   : extension revision 1
	VK_KHR_buffer_device_address          : extension revision 1
	VK_KHR_copy_commands2                 : extension revision 1
	VK_KHR_create_renderpass2             : extension revision 1
	VK_KHR_dedicated_allocation           : extension revision 3
	VK_KHR_descriptor_update_template     : extension revision 1
	VK_KHR_device_group                   : extension revision 4
	VK_KHR_draw_indirect_count            : extension revision 1
	VK_KHR_driver_properties              : extension revision 1
	VK_KHR_external_fence                 : extension revision 1
	VK_KHR_external_memory                : extension revision 1
	VK_KHR_external_semaphore             : extension revision 1
	VK_KHR_get_memory_requirements2       : extension revision 1
	VK_KHR_image_format_list              : extension revision 1
	VK_KHR_imageless_framebuffer          : extension revision 1
	VK_KHR_incremental_present            : extension revision 2
	VK_KHR_maintenance1                   : extension revision 2
	VK_KHR_maintenance2                   : extension revision 1
	VK_KHR_maintenance3                   : extension revision 1
	VK_KHR_multiview                      : extension revision 1
	VK_KHR_push_descriptor                : extension revision 2
	VK_KHR_relaxed_block_layout           : extension revision 1
	VK_KHR_sampler_mirror_clamp_to_edge   : extension revision 3
	VK_KHR_separate_depth_stencil_layouts : extension revision 1
	VK_KHR_shader_atomic_int64            : extension revision 1
	VK_KHR_shader_draw_parameters         : extension revision 1
	VK_KHR_storage_buffer_storage_class   : extension revision 1
	VK_KHR_swapchain                      : extension revision 70
	VK_KHR_uniform_buffer_standard_layout : extension revision 1
	VK_KHR_variable_pointers              : extension revision 1

VkQueueFamilyProperties:
========================
	queueProperties[0]:
	------------------
		minImageTransferGranularity = (1,1,1)
		queueCount                  = 1
		queueFlags                  = QUEUE_GRAPHICS | QUEUE_COMPUTE | QUEUE_TRANSFER
		timestampValidBits          = 64
		present support = false

VkPhysicalDeviceMemoryProperties:
=================================
memoryHeaps: count = 1
	memoryHeaps[0]:
		size   = 2147483648 (0x80000000) (2.00 GiB)
		budget = 0
		usage  = 0
		flags:
			MEMORY_HEAP_DEVICE_LOCAL_BIT
memoryTypes: count = 1
	memoryTypes[0]:
		heapIndex     = 0
		propertyFlags = 0x000f:
			MEMORY_PROPERTY_DEVICE_LOCAL_BIT
			MEMORY_PROPERTY_HOST_VISIBLE_BIT
			MEMORY_PROPERTY_HOST_COHERENT_BIT
			MEMORY_PROPERTY_HOST_CACHED_BIT
		usable for:
			IMAGE_TILING_OPTIMAL: color images, FORMAT_D16_UNORM, FORMAT_X8_D24_UNORM_PACK32, FORMAT_D32_SFLOAT, FORMAT_S8_UINT, FORMAT_D24_UNORM_S8_UINT, FORMAT_D32_SFLOAT_S8_UINT
			IMAGE_TILING_LINEAR: color images

VkPhysicalDeviceFeatures:
=========================
	robustBufferAccess                      = true
	fullDrawIndexUint32                     = true
	imageCubeArray                          = true
	independentBlend                        = true
	geometryShader                          = true
	tessellationShader                      = true
	sampleRateShading                       = true
	dualSrcBlend                            = true
	logicOp                                 = true
	multiDrawIndirect                       = true
	drawIndirectFirstInstance               = true
	depthClamp                              = true
	depthBiasClamp                          = true
	fillModeNonSolid                        = true
	depthBounds                             = false
	wideLines                               = true
	largePoints                             = true
	alphaToOne                              = true
	multiViewport                           = true
	samplerAnisotropy                       = false
	textureCompressionETC2                  = false
	textureCompressionASTC_LDR              = false
	textureCompressionBC                    = true
	occlusionQueryPrecise                   = true
	pipelineStatisticsQuery                 = true
	vertexPipelineStoresAndAtomics          = true
	fragmentStoresAndAtomics                = true
	shaderTessellationAndGeometryPointSize  = true
	shaderImageGatherExtended               = true
	shaderStorageImageExtendedFormats       = true
	shaderStorageImageMultisample           = true
	shaderStorageImageReadWithoutFormat     = false
	shaderStorageImageWriteWithoutFormat    = true
	shaderUniformBufferArrayDynamicIndexing = false
	shaderSampledImageArrayDynamicIndexing  = false
	shaderStorageBufferArrayDynamicIndexing = false
	shaderStorageImageArrayDynamicIndexing  = false
	shaderClipDistance                      = true
	shaderCullDistance                      = true
	shaderFloat64                           = true
	shaderInt64                             = true
	shaderInt16                             = true
	shaderResourceResidency                 = false
	shaderResourceMinLod                    = false
	sparseBinding                           = false
	sparseResidencyBuffer                   = false
	sparseResidencyImage2D                  = false
	sparseResidencyImage3D                  = false
	sparseResidency2Samples                 = false
	sparseResidency4Samples                 = false
	sparseResidency8Samples                 = false
	sparseResidency16Samples                = false
	sparseResidencyAliased                  = false
	variableMultisampleRate                 = false
	inheritedQueries                        = false

VkPhysicalDevice16BitStorageFeatures:
-------------------------------------
	storageBuffer16BitAccess           = true
	uniformAndStorageBuffer16BitAccess = true
	storagePushConstant16              = true
	storageInputOutput16               = false

VkPhysicalDevice8BitStorageFeaturesKHR:
---------------------------------------
	storageBuffer8BitAccess           = true
	uniformAndStorageBuffer8BitAccess = true
	storagePushConstant8              = true

VkPhysicalDeviceBufferDeviceAddressFeaturesKHR:
-----------------------------------------------
	bufferDeviceAddress              = true
	bufferDeviceAddressCaptureReplay = false
	bufferDeviceAddressMultiDevice   = false

VkPhysicalDeviceConditionalRenderingFeaturesEXT:
------------------------------------------------
	conditionalRendering          = true
	inheritedConditionalRendering = false

VkPhysicalDeviceHostQueryResetFeaturesEXT:
------------------------------------------
	hostQueryReset = true

VkPhysicalDeviceImagelessFramebufferFeaturesKHR:
------------------------------------------------
	imagelessFramebuffer = true

VkPhysicalDeviceIndexTypeUint8FeaturesEXT:
------------------------------------------
	indexTypeUint8 = true

VkPhysicalDeviceLineRasterizationFeaturesEXT:
---------------------------------------------
	rectangularLines         = true
	bresenhamLines           = true
	smoothLines              = true
	stippledRectangularLines = true
	stippledBresenhamLines   = true
	stippledSmoothLines      = true

VkPhysicalDeviceMultiviewFeatures:
----------------------------------
	multiview                   = true
	multiviewGeometryShader     = true
	multiviewTessellationShader = true

VkPhysicalDeviceProtectedMemoryFeatures:
----------------------------------------
	protectedMemory = false

VkPhysicalDeviceSamplerYcbcrConversionFeatures:
-----------------------------------------------
	samplerYcbcrConversion = false

VkPhysicalDeviceScalarBlockLayoutFeaturesEXT:
---------------------------------------------
	scalarBlockLayout = true

VkPhysicalDeviceSeparateDepthStencilLayoutsFeaturesKHR:
-------------------------------------------------------
	separateDepthStencilLayouts = true

VkPhysicalDeviceShaderAtomicInt64FeaturesKHR:
---------------------------------------------
	shaderBufferInt64Atomics = true
	shaderSharedInt64Atomics = true

VkPhysicalDeviceShaderDrawParametersFeatures:
---------------------------------------------
	shaderDrawParameters = true

VkPhysicalDeviceTransformFeedbackFeaturesEXT:
---------------------------------------------
	transformFeedback = true
	geometryStreams   = true

VkPhysicalDeviceUniformBufferStandardLayoutFeaturesKHR:
-------------------------------------------------------
	uniformBufferStandardLayout = true

VkPhysicalDeviceVariablePointersFeatures:
-----------------------------------------
	variablePointersStorageBuffer = true
	variablePointers              = false

VkPhysicalDeviceVertexAttributeDivisorFeaturesEXT:
--------------------------------------------------
	vertexAttributeInstanceRateDivisor     = true
	vertexAttributeInstanceRateZeroDivisor = false

opened by kylrth 2

如何转换成TensorRT文件 how to convert model file to tensorRT file?

我想转化成浮动尺寸推理，我按照如下步骤转换后，推理出的图像仍旧是固定尺寸。不知道问题出在哪里 1、修改难以转换成trt的网络结构。对real_esrgan模型转换，修改/basicsr/archs/rrdbnet_arch.py。将注释部分改成下边。 #feat = self.lrelu(self.conv_up1(F.interpolate(feat, scale_factor=2, mode='nearest'))) feat = self.lrelu(self.conv_up1(F.interpolate(feat, size=[int(2 * feat.shape[2]), int(2 * feat.shape[3])], mode='nearest'))) #feat = self.lrelu(self.conv_up2(F.interpolate(feat, scale_factor=2, mode='nearest'))) feat = self.lrelu(self.conv_up2(F.interpolate(feat, size=[int(2 * feat.shape[2]), int(2 * feat.shape[3])], mode='nearest'))) 2、转换成onnx 通过torch.onnx.export的参数input_names 和 output_names为onnx模型的输入和输出指定名称。一方面是防止默认名称的变化导致dynamic_axes实参错误，另外一方面也为后续转换trt文件提供便利 torch_out = torch.onnx._export(model, x, args.output, opset_version=11, export_params=True) 修改为 torch_out = torch.onnx._export(model, x, args.output, opset_version=11, export_params=True, input_names=['input'], output_names=['output'], dynamic_axes={'input':[2, 3],'output':[2,3]}) 4、转成TensorRT文件注意添加如下脚本 profile = builder.create_optimization_profile() if input_shape: profile.set_shape( # input tensor name input_name, *input_shape) config.add_optimization_profile(profile) 5、推理脚本，设置和获取shape部分 dynamic = False for i, binding in enumerate(self.engine): shape = self.engine.get_binding_shape(binding) # if input is dynamic if -1 in shape: dynamic = True self.context.active_optimization_profile = 0 if i < len(self.real_shapes): shape = self.real_shapes[i] self.context.set_binding_shape(i, shape) elif self.engine.binding_is_input(binding): raise ValueError(f'dynamic input must reset real shape') else: shape = self.context.get_binding_shape(i) logger.info('get output shape by context') elif dynamic: logger.info('output shape in trt file is fixed but input is dynamic') if i < len(self.real_shapes): shape = self.real_shapes[i] logger.info('reset output shape by real shapes') else: shape = self.context.get_binding_shape(i) logger.info('reset and get output shape by context')

opened by zhu2bowen 0
关于GPU的使用率上不去的问题 About the problem of GPU usage not increasing

运行命令:python inference_realesrgan_video.py -i inputs/01.mp4 -n realesr-animevideov3 -s 2 --suffix outx2 command:python inference_realesrgan_video.py -i inputs/01.mp4 -n realesr-animevideov3 -s 2 --suffix outx2

即使加上--num_process_per_gpu参数后,运行速度依然非常缓慢 Even after adding the --num_process_per_gpu parameter, still very slow

笔记本已经切换成独显直连模式 The laptop has been switched to the independent display direct connection mode

opened by OOPPEENN 0

Releases(v0.3.0)

v0.3.0(Sep 20, 2022)

🚀 Long time no see ☄️

✨ Highlights

✅ add realesr-general-x4v3 and realesr-general-wdn-x4v3. They are very tiny models for general scenes, and they may more robust. But as they are tiny models, their performance may be limited. ✅ support denoise strength for realesr-general-x4v3. You can use the -dn option to adjust the denising strength. Denoise strength. 0 for weak denoise (keep noise), 1 for strong denoise ability. Only used for the realesr-general-x4v3 model. ✅ update inference_video: support auto download. You do not need to download models explicitly. If the models are not download, they will be downloaded automatically.

✅ support ffmpeg stream for inference_realesrgan_video ✅ fix colorspace bug & support multi-gpu and multi-processing ✅ Added GPU selection feature to python inference ✅ deal with flv format

📢📢📢

https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-x4v3.pth

https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-general-wdn-x4v3.pth

Source code(tar.gz)
Source code(zip)
v0.2.5.0(Apr 24, 2022)
The major update: 🎉

✅ We update the RealESRGAN AnimeVideo-v3 model, which can achieve better results with a faster inference speed. The improvements are:

better naturalness

Fewer artifacts

more faithful to the original colors

better texture restoration

better background restoration

You can find more details in anime video models and comparisons.

The models can be downloaded from realesr-animevideov3.

✅ We also update the ncnn. Portable Windows / Linux / MacOS executable files for Intel/AMD/Nvidia GPU.
Source code(tar.gz)
Source code(zip)
realesr-animevideov3.pth(2.38 MB)
realesr-general-wdn-x4v3.pth(4.65 MB)
realesr-general-x4v3.pth(4.65 MB)
realesrgan-ncnn-vulkan-20220424-macos.zip(49.41 MB)
realesrgan-ncnn-vulkan-20220424-ubuntu.zip(44.75 MB)
realesrgan-ncnn-vulkan-20220424-windows.zip(43.36 MB)
v0.2.4.0(Feb 15, 2022)

Have a nice day 😸 and happy everyday 😃

I am happy to add a simple logo to Real-ESRGAN 😋 (designed by myself! and inspired by the 4K logo) So I release a new version~

You can see it on ReadMe

Source code(tar.gz)
Source code(zip)
v0.2.3.0(Dec 12, 2021)
Long long time no see😺

We have added small models for anime videos: 🎉

You can find demos and usages in https://github.com/xinntao/Real-ESRGAN/blob/master/docs/anime_video_model.md

:white_check_mark: We add small models that are optimized for anime videos :-)

Now we have added two models.

| Models | Scale | Description | | ---------------------------------------------------------------------------------------------------------------------------------- | :---- | :----------------------------- | | RealESRGANv2-animevideo-xsx2 | X2 | Anime video model with XS size | | RealESRGANv2-animevideo-xsx4 | X4 | Anime video model with XS size |

This release also contains the following files:

RealESRGANv2-animevideo-xsx2.pth

RealESRGANv2-animevideo-xsx4.pth

realesrgan-ncnn-vulkan-20211212-windows.zip

realesrgan-ncnn-vulkan-20211212-macos.zip

realesrgan-ncnn-vulkan-20211212-ubuntu.zip

Source code(tar.gz)
Source code(zip)
realesrgan-ncnn-vulkan-20211212-macos.zip(76.89 MB)
realesrgan-ncnn-vulkan-20211212-ubuntu.zip(72.29 MB)
realesrgan-ncnn-vulkan-20211212-windows.zip(71.25 MB)
RealESRGANv2-animevideo-xsx2.pth(2.30 MB)
RealESRGANv2-animevideo-xsx4.pth(2.38 MB)
v0.2.2.4(Aug 31, 2021)
See you again 😺

The major update: :tada:

:white_check_mark: We add RealESRGAN_x4plus_anime_6B.pth, which is optimized for anime images with much smaller model size. More details and comparisons with waifu2x are in anime_model.md

Other updates:

:white_check_mark: Support finetuning on your own data or paired data (i.e., finetuning ESRGAN). See here

:white_check_mark: We add CONTRIBUTING.md. Welcome to your contributions 😃

This release also contains the following files:

RealESRGAN_x4plus_anime_6B.pth

RealESRGAN_x4plus_anime_6B_netD.pth

realesrgan-ncnn-vulkan-20210901-windows.zip

realesrgan-ncnn-vulkan-20210901-macos.zip

realesrgan-ncnn-vulkan-20210901-ubuntu.zip

Source code(tar.gz)
Source code(zip)
realesrgan-ncnn-vulkan-20210901-macos.zip(74.26 MB)
realesrgan-ncnn-vulkan-20210901-ubuntu.zip(69.64 MB)
realesrgan-ncnn-vulkan-20210901-windows.zip(68.81 MB)
RealESRGAN_x4plus_anime_6B.pth(17.10 MB)
RealESRGAN_x4plus_anime_6B_netD.pth(16.77 MB)
v0.2.2.3(Aug 26, 2021)
Add finetune_realesrgan_x4plus.yml

Add scripts for data preparation:

generate_multiscale_DF2K.py

generate_meta_info.py

File list:

RealESRGAN_x4plus_netD.pth

RealESRGAN_x2plus_netD.pth

Source code(tar.gz)
Source code(zip)
RealESRGAN_x2plus_netD.pth(16.77 MB)
RealESRGAN_x4plus_netD.pth(16.77 MB)
v0.2.1(Aug 8, 2021)
Support PyPI

Support arbitrary scale with --outscale

Add RealESRGAN_x2plus.pth model

File list:

RealESRGAN_x2plus.pth

Source code(tar.gz)
Source code(zip)
RealESRGAN_x2plus.pth(63.95 MB)
v0.1.2(Jul 31, 2021)
This release is mainly for updating realesrgan-ncnn-vulkan executable files.

We have added Linux/MacOS executable files

We add back the tta option

We add the function: when the destination folder does not exist, create it!

File list:

realesrgan-ncnn-vulkan-20210801-windows.zip

realesrgan-ncnn-vulkan-20210801-macos.zip

realesrgan-ncnn-vulkan-20210801-ubuntu.zip

Source code(tar.gz)
Source code(zip)
realesrgan-ncnn-vulkan-20210801-macos.zip(95.86 MB)
realesrgan-ncnn-vulkan-20210801-ubuntu.zip(91.24 MB)
realesrgan-ncnn-vulkan-20210801-windows.zip(90.23 MB)
v0.1.1(Jul 25, 2021)
This release is mainly for storing pre-trained models and executable files.

RealESRGAN-ncnn-vulkan-20210725-windows.zip

RealESRNet_x4plus.pth

ESRGAN_SRx4_DF2KOST_official-ff704c30.pth

Note that: RealESRGAN_x4plus.pth can be found in https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth
Source code(tar.gz)
Source code(zip)
ESRGAN_SRx4_DF2KOST_official-ff704c30.pth(63.82 MB)
RealESRGAN-ncnn-vulkan-20210725-windows.zip(90.95 MB)
RealESRNet_x4plus.pth(63.93 MB)
v0.1.0(Jul 22, 2021)
This release is mainly for storing pre-trained models and executable files.

RealESRGAN-ncnn-vulkan.zip

RealESRGAN_x4plus.pth

Source code(tar.gz)
Source code(zip)
RealESRGAN-ncnn-vulkan.zip(31.78 MB)
RealESRGAN_x4plus.pth(63.93 MB)

Real-ESRGAN aims at developing Practical Algorithms for General Image Restoration.

Related tags

Overview

Real-ESRGAN

📖 Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data

Portable executable files

🔧 Dependencies and Installation

Installation

⚡ Quick Inference

🏰 Model Zoo

💻 Training

BibTeX

📧 Contact

Comments

Releases(v0.3.0)

v0.3.0(Sep 20, 2022)

v0.2.5.0(Apr 24, 2022)

v0.2.4.0(Feb 15, 2022)

v0.2.3.0(Dec 12, 2021)

v0.2.2.4(Aug 31, 2021)

v0.2.2.3(Aug 26, 2021)

v0.2.1(Aug 8, 2021)

v0.1.2(Jul 31, 2021)

v0.1.1(Jul 25, 2021)

v0.1.0(Jul 22, 2021)

Owner

Xintao

PyTorch implementation of a Real-ESRGAN model trained on custom dataset

My usage of Real-ESRGAN to upscale anime, some test and results in the test_img folder

An official repository for Paper "Uformer: A General U-Shaped Transformer for Image Restoration".

Official repository for "Restormer: Efficient Transformer for High-Resolution Image Restoration". SOTA for motion deblurring, image deraining, denoising (Gaussian/real data), and defocus deblurring.

Scripts of Machine Learning Algorithms from Scratch. Implementations of machine learning models and algorithms using nothing but NumPy with a focus on accessibility. Aims to cover everything from basic to advance.

This project is based on RIFE and aims to make RIFE more practical for users by adding various features and design new models

A toolkit for developing and comparing reinforcement learning algorithms.

FedJAX is a library for developing custom Federated Learning (FL) algorithms in JAX.

Machine Learning From Scratch. Bare bones NumPy implementations of machine learning models and algorithms with a focus on accessibility. Aims to cover everything from linear regression to deep learning.

All the essential resources and template code needed to understand and practice data structures and algorithms in python with few small projects to demonstrate their practical application.

Multi-Stage Progressive Image Restoration

(under submission) Bayesian Integration of a Generative Prior for Image Restoration

HINet: Half Instance Normalization Network for Image Restoration

This is an implementation for the CVPR2020 paper "Learning Invariant Representation for Unsupervised Image Restoration"

EDPN: Enhanced Deep Pyramid Network for Blurry Image Restoration

Image restoration with neural networks but without learning.

SwinIR: Image Restoration Using Swin Transformer

Image Restoration Toolbox (PyTorch). Training and testing codes for DPIR, USRNet, DnCNN, FFDNet, SRMD, DPSR, BSRGAN, SwinIR

Dynamic Attentive Graph Learning for Image Restoration, ICCV2021 [PyTorch Code]