profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/xinntao/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Xintao xinntao Tencent Shenzhen, China https://xinntao.github.io/ Researcher at Tencent ARC Lab, (Applied Research Center)

xinntao/ESRGAN 3705

ECCV18 Workshops - Enhanced SRGAN. Champion PIRM Challenge on Perceptual Super-Resolution. The training codes are in BasicSR.

TencentARC/GFPGAN 3418

GFPGAN aims at developing Practical Algorithms for Real-world Face Restoration.

open-mmlab/mmediting 2390

OpenMMLab Image and Video Editing Toolbox

xinntao/BasicSR 2328

Open Source Image and Video Restoration Toolbox for Super-resolution, Denoise, Deblurring, etc. Currently, it includes EDSR, RCAN, SRResNet, SRGAN, ESRGAN, EDVR, BasicVSR, SwinIR, etc. Also support StyleGAN2, DFDNet.

xinntao/EDVR 1232

Winning Solution in NTIRE19 Challenges on Video Restoration and Enhancement (CVPR19 Workshops) - Video Restoration with Enhanced Deformable Convolutional Networks. EDVR has been merged into BasicSR and this repo is a mirror of BasicSR.

xinntao/DNI 104

CVPR19 - Deep Network Interpolation for Continuous Imagery Effect Transition

xinntao/HandyView 90

Handy image viewer based on PyQt5. Convenient for viewing and comparing :-)

xinntao/facexlib 64

FaceXlib aims at providing ready-to-use face-related functions based on current STOA open-source methods.

xinntao/BasicSR-examples 13

BasicSR-Examples illustrates how to easily use BasicSR in your own project

issue commentTencentARC/GFPGAN

Is this possible, to improve?

@zelenooki87 Thanks:-)

So many family photos. You have a warm and loving family 😃

zelenooki87

comment created time in 29 minutes

issue commentTencentARC/GFPGAN

ImportError: No module named 'deform_conv'

@ntquyen11 What env do you use? and you may need to put more information/error log here

EndeCC0

comment created time in 32 minutes

issue commentxinntao/Real-ESRGAN

no vulkan device/invalid gpu device

这个提示, 推测应该是vulkan device 没有办法使用。 我对这块不熟悉。 你可以在ncnn中尝试看看能否得到解决:https://github.com/Tencent/ncnn 里面有技术交流群~

DingKuiqing

comment created time in 36 minutes

issue commentxinntao/Real-ESRGAN

怎麽能在Cupscale GUI裏調用Real-ESRGAN的pytorch AI網絡?

我理解 Cupscale GUI 只能使用ncnn来做, 不适合 直接使用pytorch。

如果需要直接使用pytorch,还是需要安装环境,然后命令行中使用。 具体如何解决,我也不是太清楚~

Battlecraft369

comment created time in 38 minutes

issue commentTencentARC/GFPGAN

How to get original size? (no upscale)

--upscale 1 should be work 😄

I will close this issue. If you still have any questions, please re-open it. Thanks~

AlonDan

comment created time in 40 minutes

issue commentxinntao/Real-ESRGAN

will it work on text images?

It will work on text images. But you need to finetune on those images with texts.

YC7225

comment created time in 42 minutes

issue commentTencentARC/GFPGAN

Have the same question as "Training with clean version #61"

It will take more time to add the detailed training process for GFPGAN-clean and all those pre-trained models.

Maybe in November. Thanks

linkboyx

comment created time in 44 minutes

pull request commentTencentARC/GFPGAN

Add Docker environment & web demo

@CJWBW Thanks for your work 😄

I wonder that could Replicate use another branch instead of the master branch?

As it will introduce some files (like download-weights, predict.py and run_setup.sh). These files are specifically for running on Replicate, and users may confuse with those files when they run on their local machines.

A better solution is to use another branch (for example replicate), and in that branch, we add those files and then ‘deploy’ on the replicate website.

CJWBW

comment created time in an hour

issue commentTencentARC/GFPGAN

About training with 8 gpus

  1. I have also trained with 8 GPUs without modifying the hyper-parameters. The results are Ok with fewer iterations. But it may be not the optimal~
NNNNAI

comment created time in an hour

issue commentxinntao/Real-ESRGAN

Support on grayscale image and single GPU train from scratch

  1. The trained model takes the images with three channels. But the inference codes can deal with gray images with 8 bit/16bit (actually by repeating the gray channels to three channels). See more details here: https://github.com/xinntao/Real-ESRGAN/blob/90ddf13b5eb11c9c9f83c944078d968f9dcbdf8f/realesrgan/utils.py#L147-L165

You can try your own gray images. If you find something wrong, please let us know. Thanks:-)

  1. We add training commands on a single GPU: https://github.com/xinntao/Real-ESRGAN/blob/master/Training.md#Train-Real-ESRNet For example: python realesrgan/train.py -opt options/train_realesrgan_x4plus.yml --auto_resume
kail85

comment created time in an hour

push eventxinntao/Real-ESRGAN

Xintao

commit sha aa584e05bcc8eed89cf4a9835a842f6c9918f33c

minor updates on Training.md

view details

push time in an hour

push eventxinntao/Real-ESRGAN

Xintao

commit sha b525d1793beddc6231f82bee2ce1a577a69559ee

add trainining with one gpu

view details

push time in an hour

push eventxinntao/Real-ESRGAN

Xintao

commit sha 0ad2e9c61e158883353ebf99ff5392e31d760c8d

set num_gpu to auto in options

view details

push time in an hour

issue commentxinntao/Real-ESRGAN

Is it possible that ".jftf" is not supported?

  1. you can try to save to jpg images, by adding this argument in your commands: --ext jpg
  2. If you want to save to images with the extension.jftf, it may require other packages.
NaokiSato102

comment created time in 2 hours

delete branch xinntao/BasicSR

delete branch : metric

delete time in 16 hours

push eventxinntao/BasicSR

Xintao

commit sha a9bf05a426e10e4a97fdc206c7324ac36920e07f

Support multiple inputs for metrics during validation (#467) * be compatible with niqe * convert niqe to float type * update path for niqe_pris_params

view details

push time in 16 hours

push eventxinntao/BasicSR

Xintao

commit sha c6f3ec729cf1f0921ff259d1e758de2421dcef20

update path for niqe_pris_params

view details

push time in 16 hours

push eventxinntao/BasicSR-examples

Xintao

commit sha c017e31eb3b670eabe6625a613d67f33a8540588

add calculate_niqe option

view details

Xintao

commit sha b644b2575532e2580890f19e4ba67544d136b955

Merge branch 'master' of github.com:xinntao/BasicSR-examples into installation

view details

push time in 16 hours

push eventxinntao/BasicSR-examples

Xintao

commit sha c017e31eb3b670eabe6625a613d67f33a8540588

add calculate_niqe option

view details

push time in 16 hours

PR opened xinntao/BasicSR

Model supports multiple inputs for metrics during validation
+36 -29

0 comment

5 changed files

pr created time in 18 hours

push eventxinntao/BasicSR

Xintao

commit sha 315156a227ceb2d67f845dc711b244b312e6fd3b

convert niqe to float type

view details

push time in 18 hours

create barnchxinntao/BasicSR

branch : metric

created branch time in 18 hours

push eventxinntao/BasicSR

Xintao

commit sha 0830e4645ac33f3a040d85cf1a363ac987b0437a

use logger.warning instead of logger.warn

view details

push time in 4 days

release xinntao/BasicSR

v1.3.4.2

released time in 4 days

created tagxinntao/BasicSR

tagv1.3.4.2

Open Source Image and Video Restoration Toolbox for Super-resolution, Denoise, Deblurring, etc. Currently, it includes EDSR, RCAN, SRResNet, SRGAN, ESRGAN, EDVR, BasicVSR, SwinIR, etc. Also support StyleGAN2, DFDNet.

created time in 4 days

push eventxinntao/BasicSR

Xintao

commit sha 4a75f5b7298645f154a9e8463e59ad52ceecde63

add vscode settings; V1.3.4.2

view details

push time in 4 days

issue commentTencentARC/GFPGAN

Is this possible, to improve?

Thanks 👍

Your suggestion is good. I updated a version of facexlib, which removes too small faces or side faces, for failure cases as you provided. More details: https://github.com/xinntao/facexlib/commit/24493fc888acb03ca627ed07cd75376ab56c85bc

But it may be not perfect, as it now hard-codes the eye-to-eye distance as a threshold. I will update it later, and will also consider your suggestion that modify code if face area is too small or blur that then model not generate output.

Thanks

zelenooki87

comment created time in 7 days

push eventxinntao/facexlib

push time in 7 days

issue commentxinntao/Real-ESRGAN

Wired output

@iBobbyTS You can now update facexlib to 0.2.1.0 for fix this bug: https://pypi.org/project/facexlib/0.2.1.0/

iBobbyTS

comment created time in 7 days