AnimeGANv2 photo animation using ModelArts

前言

---

ModelArts 是面向开发者的一站式 AI 开发平台,为机器学习与深度学习提供海量数据预处理及交互式智能标注、大规模分布式训练、自动化模型生成,及端-边-云模型按需部署能力,帮助用户快速创建和部署模型,管理全周期 AI 工作流.

前期准备

---

在使用ModelArts之前,需要进入华为云官网 https://www.huaweicloud.com/ ,然后注册华为云账号,再进行实名认证.主要分为3步（注册–>实名认证–>服务授权）（如有已完成部分,请忽略）

点此去完成 实名认证,账号类型选"个人",个人认证类型推荐使用"扫码认证".

点此进入 ModelArts 控制台数据管理页面,上方会提示访问授权,点击【服务授权】按钮,按下图顺序操作：

具体步骤

---

The following is the concrete steps of the photos anime

第一步 进入项目

Enter the huawei cloud provided by AnimeGANv2照片动漫化 项目入口：https://developer.huaweicloud.com/develop/aigallery/notebook/detail?id=b2f6ba98-34d9-4f0a-b5b2-35133fe2421c,点击 Run in ModelArts.

Just went in displays connection,And the initialization of the,Wait to initialize the end.

点击 切换规格.选择 [限时免费]GPU: 1*P100|CPU: 8核 64GB,点击切换规格.等待切换.

切换完毕,点击 确定.

切换kernel为PyTorch-1.4,点击Select切换到文件夹,The directory file structure is as follows,ipynbRun path for "/home/ma-user/work/ma_share/AnimeGanv2/AnimeGanv2.ipynb"第二步 Get the code and data

选中以下代码,点击左侧运行按钮,Or the cursor to where the codecell,快捷键Ctrl + Enter 运行代码.

First run please execute the code,The first run do not perform this code again,Path data repetition and confusion

运行结果：

       

This project required data downloading,刷新路径,Can display the data document

     

第三步 安装依赖库

Click the run button orCtrl + Enter Run the following two pieces of code,To rely on the installation of the library.The running time is a bit long,请耐心等待.

部分运行结果如下,Said successful installation dependent libraries.   

%cd AnimeGANv2  # 进入数据目录,Image upload please refer this path# /home/ma-user/work/AnimeGANv2

第四步 Generate widely in

Run the subsequent three code（部分截图）.

运行下面代码,Can show the project the resultsSelect the project in other local pictures or upload pictures,Replace the code use cases in the picture,并运行.如出现FileNotFoundError错误时,Please check the file storage path is wrong,Or replace with absolute path case pictures

After the operation generated animation picture：

Drag and drop the generated cartoon pictures to local directory,即可保存.

Run the following two pieces of code,即可对urlImage links to generate widely in.Can modify the pictureurl地址进行测试

体验总结

---

It is smooth all the way down,Midway installation dependent libraries may have to wait some time,尤其是要 切换规格 The default specification may install failure.简言之,ModelArts非常适合零基础入门AI的同学,There will be a great experience！

参考文献

---

[1] https://developer.huaweicloud.com/develop/aigallery/notebook/detail?id=b2f6ba98-34d9-4f0a-b5b2-35133fe2421c