张小白AI博文知识点串烧Ver0.1-20210422
ModelArts也好,MindSpore也好,昇腾200DK,CANN也好,很多知识点交叉在一起,张小白有时候都觉得自己写的博客或论坛帖子中,有很多重复的地方。想想自己应该稍微整理下,故形成此文,供自己和有心的人参考。
(零)一些知识点:
Ubuntu
系统安装
(1)安装Ubuntu 18.04.1 X86虚拟机 https://bbs.huaweicloud.com/blogs/198324 (Part Two ) https://bbs.huaweicloud.com/blogs/198349 (Part Three)
(2)安装Ubuntu 18.04 Server后再安装desktop桌面(解决显卡冲突问题) https://bbs.huaweicloud.com/blogs/198357
(3)安装18.04 Desktop:https://bbs.huaweicloud.com/blogs/194291 (Part II)
软件安装
(1)Ubuntu 18.04更新apt源 https://bbs.huaweicloud.com/blogs/198349 (搜索 替换apt源)
(2)Git安装 https://bbs.huaweicloud.com/blogs/195227 (搜索 安装个git)
(3)Ubuntu18.04 安装docker https://bbs.huaweicloud.com/blogs/241465
(4)Ubuntu RSTP视频流live555 https://bbs.huaweicloud.com/blogs/195833 (搜索 使用live555制作rtsp视频流 )结合 VLC media player
编译相关
(1)Ubuntu 18.04源码安装gcc 7.3.0 https://bbs.huaweicloud.com/blogs/198349 (Part Two)
(2)Ubuntu 安装cmake https://bbs.huaweicloud.com/blogs/198349 (搜索 cmake 4.1)
Python 3.7.5
(1)Ubuntu 18.04 Python 3.7.5 安装 https://bbs.huaweicloud.com/blogs/195227 (搜索 python的源码包) https://bbs.huaweicloud.com/blogs/198324 (Part Three)
(2)Ubuntu 18.04 docker安装Python 3.7.5 https://bbs.huaweicloud.com/blogs/241465 (搜索 拉取python的docker镜像)
显卡驱动
(1)Ubuntu 18.04裸机安装CUDA 10.1 https://bbs.huaweicloud.com/blogs/198357 (搜索 装CUDA)
(2)Ubuntu 18.04裸机安装cuDNN 7.6.5 https://bbs.huaweicloud.com/blogs/198357 (搜索 cuDNN )
Windows
软件安装
(1)JupyterLab安装 https://bbs.huaweicloud.com/blogs/260796 (搜索 安装Jupyter Lab for windows )
编译相关
(1)Win10 MinGW GCC安装 https://bbs.huaweicloud.com/blogs/239829 (搜索 MinGW GCC )
(2)Win10 Cmake 3.19.3安装 https://bbs.huaweicloud.com/blogs/239829 (搜索 Windows版CMake)
显卡驱动
(1)Win10 GPU驱动CUDA和cuDNN安装 https://bbs.huaweicloud.com/blogs/212446
Ubuntu ARM
(1)ffmpeg ARM版本安装 https://bbs.huaweicloud.com/blogs/195391 (搜索 ffmpeg )
(2)opencv ARM版本安装 https://bbs.huaweicloud.com/blogs/195391 (搜索 安装opencv )
Android Studio
(1)Android Studio 安装NDK和CMake等SDK https://bbs.huaweicloud.com/blogs/205515 (搜索 SDK)
(2)Android Studio解决apk签名问题 https://bbs.huaweicloud.com/blogs/205515 (搜索 带签名的安装包 )
工具与技巧
(1)ssh服务端和客户端SecureCRT,scp文件传输SecureFX的安装使用:https://bbs.huaweicloud.com/blogs/195227 (搜索 scp传输 )
(2)视频转换为图片: https://bbs.huaweicloud.com/blogs/207008 (搜索 视频转换为图片)
(3)图片合成视频:https://bbs.huaweicloud.com/blogs/207008 (搜索 合成视频)
(4)Github无法连接问题的变通 https://bbs.huaweicloud.com/blogs/207322 (搜索 github不是你想连就能连 )
(一)Atlas 200DK:
(1)200DK开机体验 https://bbs.huaweicloud.com/blogs/193847
16.04版本
(1)使用ADK Installer工具制作 16.04 ARM SD系统盘 https://bbs.huaweicloud.com/blogs/193847(搜索 ADKInstaller)
(2)安装和配置MindStudio 2.1 (16.04配套)https://bbs.huaweicloud.com/blogs/193847 (搜索 在MindStudio上安装和配置DDK )
(3)编译第一个应用: https://bbs.huaweicloud.com/blogs/193847(搜索 创建首个AI应用 )
(4)摄像头: https://bbs.huaweicloud.com/blogs/194291 (sample-ascendcamera 摄像头采集图片)
(5)视频采集: https://bbs.huaweicloud.com/blogs/194641
(6)YoloV3目标检测: https://bbs.huaweicloud.com/blogs/194816 (使用fast-deploy部署 sample-objectdetectionbyyolov3 )
(7)人体检测 https://bbs.huaweicloud.com/blogs/195833 (使用fast-deploy部署 sample-videoanalysisperson )
18.04版本:
(1)交叉编译环境安装: https://bbs.huaweicloud.com/blogs/195227 (搜索 两个ToolKit)
(2)安装MindStudio 2.3.3(18.04配套) https://bbs.huaweicloud.com/blogs/195227(搜索 下载2.3.3版本的MindStudio )
(3)开发板制卡:https://bbs.huaweicloud.com/blogs/195227 (搜索 开发板的制卡 )
(4)连接200DK网卡配置 https://bbs.huaweicloud.com/blogs/195227 (搜索 虚拟网卡ens160u4u1 )
(5)使用MindStudio 2.3.3 进行模型转换 https://bbs.huaweicloud.com/blogs/195227 (搜索 下载权重文件 )https://bbs.huaweicloud.com/blogs/195391 (搜索 转换模型 )
(6)ResNet样例程序 https://bbs.huaweicloud.com/blogs/195227 (搜索 转换好的模型导入到工程 )
(7)黑白图像上色 https://bbs.huaweicloud.com/blogs/195391 (运行环境Ubuntu ARM系统编译,开发环境Ubuntu X86系统MindStudio远程运行 )
(8)黑白视频上色 https://bbs.huaweicloud.com/blogs/195590 ( 开发环境交叉编译第三方库,导入运行环境。运行环境本地执行)
CANN 3.3.0alpha1
(1)200DK安装 CANN 3.3.0alpha https://bbs.huaweicloud.com/forum/forum.php?mod=viewthread&tid=116719
其他:
(1)验证16.04与18.04切换 https://bbs.huaweicloud.com/blogs/195227 (搜索 将18.04的sd卡换回16.04的sd )
(2)200DK的系统备份 https://bbs.huaweicloud.com/blogs/196033
(二)HiLens
(1)口罩检测 https://bbs.huaweicloud.com/blogs/194916
(三)MindSpore
安装
MindSpore
(1)Ubuntu 18.04 MindSpore V1.0 CPU版本PIP安装 https://bbs.huaweicloud.com/blogs/198324
(2)Ubuntu 18.04 MindSpore V1.0 CPU版本源码安装 https://bbs.huaweicloud.com/blogs/198349
(3)Ubuntu 18.04 MindSpore V1.0 GPU版本PIP安装 https://bbs.huaweicloud.com/blogs/198357
(4)ubuntu 18.04 MindSpore V1.1.1 docker+PIP安装 https://bbs.huaweicloud.com/blogs/241465
(5)MacOS安装 MindSpore V1.2.0-rc https://bbs.huaweicloud.com/blogs/255527
MindSpore Lite
(1)Win10 MindSpore Lite V1.1.0 CPU版本源码安装 https://bbs.huaweicloud.com/blogs/239829
(2)Ubuntu 18.04 MindSpore Lite V1.1.0 CPU版本源码安装 https://bbs.huaweicloud.com/blogs/239830
特性
(1)PyNative模式 https://bbs.huaweicloud.com/blogs/242300
(2)单节点缓存 CPU: https://bbs.huaweicloud.com/blogs/242306 GPU: https://bbs.huaweicloud.com/blogs/242307
(3)端侧推理 https://bbs.huaweicloud.com/blogs/242317
神经网络训练与推理
(1)MindSpore LENET网络训练 https://bbs.huaweicloud.com/blogs/198324 (Part Five)https://bbs.huaweicloud.com/blogs/198349 (Part Seven) https://bbs.huaweicloud.com/blogs/242300 (PyNative模式)https://bbs.huaweicloud.com/blogs/260796(JupyterLab体验)
(2)Bert中文新闻分类(ModelArts) https://bbs.huaweicloud.com/blogs/207332
(3)Resnet50毒蘑菇识别(ModelArts)https://bbs.huaweicloud.com/blogs/208652
(4)YoloV3-DarkNet50篮球检测(ModelArts)https://bbs.huaweicloud.com/blogs/207008
(5)Wide&Deep CTR预估(ModelArts,PyCharm Toolkit)https://bbs.huaweicloud.com/blogs/207322
(6)ResNet cifar10 (Cache训练)https://bbs.huaweicloud.com/blogs/242307
(7)使用Android Studio实现图像分类 https://bbs.huaweicloud.com/blogs/205515
(8)使用Android Studio实现图像检测 (object_detection) https://bbs.huaweicloud.com/blogs/208653
模型转换
(1)使用 converter_lite转换mindir模型为ms模型 https://bbs.huaweicloud.com/blogs/239829 (搜索 converter_lite.exe )https://bbs.huaweicloud.com/blogs/239830 (搜索 尝试converter )
(四)ModelArts
(1)ModelArts的PyCharm Kit https://bbs.huaweicloud.com/blogs/207322 (三、PyCharm Kit的登场)
(2)自动学习 毒蘑菇识别 https://bbs.huaweicloud.com/blogs/208172
(3)蚂蚁牙黑 ModelArts实现 https://bbs.huaweicloud.com/blogs/246301 本地实现 https://bbs.huaweicloud.com/blogs/246302
(五)AI云服务器
(1)体验CANN 3.3.0 https://bbs.huaweicloud.com/blogs/254154
张小白将持续完善此内容,以便今后的检索需要。
(全文完,谢谢阅读)
- 点赞
- 收藏
- 关注作者
评论(0)