当前位置:网站首页>Image and video restoration and super score - use of real esrgan project (I) | machine learning
Image and video restoration and super score - use of real esrgan project (I) | machine learning
2022-07-21 18:05:00 【Swordsman a Liang_ ALiang】
Catalog
Picture test repair super score
Preface
I've been doing it some time ago golang, Pay little attention to some open source projects . Happened to meet one , You can make blurred photos or videos clear , Projects that can be processed in excess .
project github Address :GitHub - xinntao/Real-ESRGAN: Real-ESRGAN aims at developing Practical Algorithms for General Image/Video Restoration.
Take a look at the renderings given by the project .
It can be seen that the model also has a good effect on animation illustration .
The project also lists some similar projects , You can see GFPGAN It's on the list .
If you're right GFPGAN If you are also interested , You can refer to another article of mine to understand .
This article mainly makes some simple use of this project .
Download the execution file
You can directly download the execution program of the project to run and see .
If you can't download , You can use my link below .
link :https://pan.baidu.com/s/1Vqzp4ylIf-YB-Bw0PVrYog
Extraction code :tuan
Test use
The unzipped directory structure is shown in the figure below .
We can go through README file , Learn how to use .
Picture test repair super score
Let's take a look at the pictures provided and their resolution .
open CMD Go to the folder location .
Execute the order as follows :
realesrgan-ncnn-vulkan.exe -i input.jpg -o output.png -n realesr-animevideov3
The results are as follows , You can see the specific progress .
Finally, the processed image is generated output.png, Let's take a look at the generated image and its resolution .
It can be seen that both the definition and resolution have been improved .
Video test repair super score
Video needs to be used ffmpeg Split the video , If not ffmpeg Of , Please refer to my article .
First create two directories to store split and processed picture frames .
First, let's take a look at the provided video picture and its resolution .
according to README The instructions in it , There are three steps to deal with .
step1
ffmpeg -i onepiece_demo.mp4 -qscale:v 1 -qmin 1 -qmax 1 -vsync 0 tmp_frames/frame%08d.jpg
After execution tmp_frames The directory produces split picture frames .
step2
realesrgan-ncnn-vulkan.exe -i tmp_frames -o out_frames -n realesr-animevideov3 -s 2 -f jpg
This step will take a long time .
step3
ffmpeg -i out_frames/frame%08d.jpg -i onepiece_demo.mp4 -map 0:v:0 -map 1:a:0 -c:a copy -c:v libx264 -r 23.98 -pix_fmt yuv420p output_w_audio.mp4
take out_frames The pictures in are compressed into video , Take a look at the final video quality .
It can be seen that the definition and resolution have also been greatly improved .
summary
What attracts me more about this project is its support for animation illustrations , There are special models .
Recently, I am learning the first principle , I thought it over . In fact, we have two methods, induction and deduction , Just subconsciously choose the first . Still stand out of things , Adjust your way of thinking , To have more vision .
边栏推荐
- QDU暑假集训第一周限时训练1
- vim 编辑器学习笔记
- Jianghu nickname of Chinese Universities
- What is the situation of foreign lead industry? Mentality collapse until giving up
- Talk about a wave of moveit2
- Use the esp32 download tool to burn firmware for esp32
- 架构实战营模块7作业
- Explain cache penetration, cache avalanche, cache breakdown
- Bit operation of code
- Meaning of downstream task
猜你喜欢
Raspberry pie 3B builds Flink cluster
Clickhouse CPU memory resource optimization configuration
Homeland defense | technical guide for whole stack IOT irrigation system
[CS231N]Notes_ 1-Introduction
Understand RTK Positioning, this one is enough!
The most complete summary of MySQL data types in history - (first)
leetcode:1125. 最小的必要团队【状压dp板子 + 集合覆盖板子】
Installation and configuration of CodeBlocks
About web responsive design
20220720学习反思
随机推荐
Kali carries out ARP network disconnection attack
聊聊Redis内存优化的7个神技
Video 4413.8 regional convolutional neural network (r-cnn) series 13.5 multiscale target detection & 13.6 target detection data set
Vidéo d'utilisation à deux points 41 13.3. Détection des cibles et cadre de limitation AQ 13.4 cadre d'ancrage
史上最全的mysql数据类型汇总-(上)
Talk about a wave of moveit2
力扣139题:单词拆分
What is the situation of foreign lead industry? Mentality collapse until giving up
212. 单词搜索 II
知识图谱赋能数字经济,第十六届全国知识图谱与语义计算大会(CCKS 2022)即将召开
Colon usage video 41 13.3 Target detection and bounding box QA 13.4 anchor box
leetcode:1125. 最小的必要团队【状压dp板子 + 集合覆盖板子】
小程序的破局之道,数字化营销已然成为趋势
02_ UE4 advanced_ HP strip and blood deduction mechanism
Vulnerability statements of the sorted burp official website
The new version of hands-on learning ros2 has been released
Three ways of cache mapping
从日常小事看如何工作
Super milk's note taking software: obsidain
越来越多的场景和行业开始与元宇宙扯上关系