一、基礎(chǔ)環(huán)境

硬件：KV260視覺入門套件

攝像頭：?？?20P USB攝像頭（因?yàn)椴渴餠itis AI之后懶得poweroff插MIPI camera了，直接USB上）

軟件：Ubuntu 22.04 + Vitis AI v3.0

root@kria:~# uname -a
Linux kria 5.15.0-1020-xilinx-zynqmp #22-Ubuntu SMP Fri Feb 2414:14:20 UTC 2023 aarch64 aarch64 aarch64 GNU/Linux
root@kria:~#

二、初步體驗(yàn)

Vitis AI 3.0的用戶手冊(cè)如下，基本概念和指南很清晰，只需要一步一步照著做即可。

Vitis AI — Vitis? AI 3.0 documentation (xilinx.github.io)

先拿smartcam來開刀，展示下具體流程，先apt update/upgrade全部走一遍。

（一）Docker安裝&配置

Vitis AI v3.0通過Docker部署，故需要安裝&配置Docker環(huán)境。
apt install docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin

隨后驗(yàn)證安裝是否成功：

因?yàn)橐恢眗oot，所以可以忽略Docker用戶權(quán)限。

之后添加國(guó)內(nèi)Docker鏡像源，否則2.4G的smartcam鏡像一晚上根本下不完。

先在/etc/docker/ 目錄下增加daemon.json文件，將國(guó)內(nèi)鏡像源加進(jìn)去，文件內(nèi)容如下：

{
    "registry-mirrors": [
        "https://ccr.ccs.tencentyun.com",
        "https://docker.mirrors.ustc.edu.cn",
        "https://hub-mirror.c.163.com",
        "https://mirror.baidubce.com",
        "https://registry.docker-cn.com"
    ]
}

之后service docker restart重啟Docker服務(wù)，再docker info檢查是否OK，差不多就是下面的樣子：

Cgroup Version: 2
 Plugins:
  Volume: local
  Network: bridge host ipvlan macvlan null overlay
  Log: awslogs fluentd gcplogs gelf journald json-file local logentries splunk syslog
 Swarm: inactive
 Runtimes: io.containerd.runc.v2 runc
 Default Runtime: runc
 Init Binary: docker-init
 containerd version: 8165feabfdfe38c65b599c4993d227328c231fca
 runc version: v1.1.8-0-g82f18fe
 init version: de40ad0
 Security Options:
  apparmor
  seccomp
   Profile: builtin
  cgroupns
 Kernel Version: 5.15.0-1023-xilinx-zynqmp
 Operating System: Ubuntu 22.04.3 LTS
 OSType: linux
 Architecture: aarch64
 CPUs: 4
 Total Memory: 3.814GiB
 Name: kria
 ID: d0241500-7aff-4090-bba3-e961db7c1818
 Docker Root Dir: /var/lib/docker
 Debug Mode: false
 Experimental: false
 Insecure Registries:
  127.0.0.0/8
 Registry Mirrors:
  https://ccr.ccs.tencentyun.com/
  https://docker.mirrors.ustc.edu.cn/
  https://hub-mirror.c.163.com/
  https://mirror.baidubce.com/
  https://registry.docker-cn.com/
 Live Restore Enabled: false

（二）配置Vitis AI

先添加Xilinx軟件包源，通過add-apt-repository ppa:xilinx-apps/ppa和add-apt-repository ppa:ubuntu-xilinx/updates兩個(gè)cmd設(shè)置額外的apt repository。

之后，apt update更新所有package。

root@kria:~/Vitis-AI# apt update
Hit:1 https://download.docker.com/linux/ubuntu jammy InRelease
Hit:2 http://ports.ubuntu.com/ubuntu-ports jammy InRelease
Hit:3 http://ports.ubuntu.com/ubuntu-ports jammy-updates InRelease
Hit:4 http://oem.archive.canonical.com/updates jammy-limerick InRelease
Hit:5 https://ppa.launchpadcontent.net/ubuntu-xilinx/sdk/ubuntu jammy InRelease
Hit:6 http://ports.ubuntu.com/ubuntu-ports jammy-backports InRelease
Hit:7 http://ports.ubuntu.com/ubuntu-ports jammy-security InRelease
Hit:8 https://ppa.launchpadcontent.net/ubuntu-xilinx/updates/ubuntu jammy InRelease
Hit:9 https://ppa.launchpadcontent.net/xilinx-apps/ppa/ubuntu jammy InRelease
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
5 packages can be upgraded. Run 'apt list --upgradable' to see them.
root@kria:~/Vitis-AI# apt upgrade
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
Calculating upgrade... Done
Get more security updates through Ubuntu Pro with 'esm-apps' enabled:
  python2.7-minimal libjs-jquery-ui libopenexr25 libavcodec58 libavutil56
  libswscale5 libswresample3 libavformat58 python2.7 libpython2.7-minimal
  libpython2.7-stdlib
Learn more about Ubuntu Pro at https://ubuntu.com/pro
The following packages have been kept back:
  gjs libgjs0g libsmbclient libwbclient0 samba-libs
0 upgraded, 0 newly installed, 0 to remove and 5 not upgraded.

后續(xù)，繼續(xù)apt install xlnx-firmware-kv260-smartcam下載smartcam固件。

root@kria:~/Vitis-AI# apt install xlnx-firmware-kv260-smartcam
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
The following NEW packages will be installed:
  xlnx-firmware-kv260-smartcam
0 upgraded, 1 newly installed, 0 to remove and 5 not upgraded.
Need to get 2973 kB of archives.
After this operation, 7921 kB of additional disk space will be used.
Get:1 https://ppa.launchpadcontent.net/xilinx-apps/ppa/ubuntu jammy/main arm64 xlnx-firmware-kv260-smartcam arm64 0.9-0xlnx1 [2973 kB]
Fetched 2973 kB in 4s (727 kB/s)
Scanning processes...
Restarting services...ocessor microcode upgrades.
Package configuration

通過xmutil命令xmutil listapps查看：

Active_slot為0的表示正在運(yùn)行，可以xmutil unloadapp停止掉，這之前還要xmutil desktop_disable禁用桌面，奇怪的是xmutil unloadapp運(yùn)行后，風(fēng)扇狂轉(zhuǎn)，按理來說停止服務(wù)應(yīng)該資源消耗變小了，可是fan那么激動(dòng)做啥？

還有個(gè)提醒memory leak的ERROR，有知道的大佬嗎？

root@kria:~/Vitis-AI# xmutil unloadapp
[ 8191.766515] OF: ERROR: memory leak, expected refcount 1 instead of 2, of_node_get()/of_node_put() unbalanced - destroy cset entry: attach overlay node /axi/zyxclmm_drm
[ 8191.781699] OF: ERROR: memory leak before free overlay changeset,  /axi/isp_vcap_csi/ports/port@0/endpoint
[ 8191.793445] OF: ERROR: memory leak, expected refcount 1 instead of -1073741824, of_node_get()/of_node_put() unbalanced - destroy cset entry: attach overlay node /axi/isp_vcap_csi/ports/port@0/endpoint
[ 8191.811821] OF: ERROR: memory leak before free overlay changeset,  /axi/scaler@b0100000/ports/port@1/endpoint
[ 8191.821849] OF: ERROR: memory leak, expected refcount 1 instead of -1073741824, of_node_get()/of_node_put() unbalanced - destroy cset entry: attach overlay node /axi/scaler@b0100000/ports/port@1/endpoint
[ 8191.840101] OF: ERROR: memory leak before free overlay changeset,  /axi/scaler@b0100000/ports/port@0/endpoint
[ 8191.850111] OF: ERROR: memory leak, expected refcount 1 instead of -1073741824, of_node_get()/of_node_put() unbalanced - destroy cset entry: attach overlay node /axi/scaler@b0100000/ports/port@0/endpoint
[ 8191.868361] OF: ERROR: memory leak, expected refcount 1 instead of 2, of_node_get()/of_node_put() unbalanced - destroy cset entry: attach overlay node /axi/scaler@b0100000/ports
[ 8191.884328] OF: ERROR: memory leak before free overlay changeset,  /axi/csiss@80000000/ports/port@1/endpoint
[ 8191.894253] OF: ERROR: memory leak, expected refcount 1 instead of -1073741824, of_node_get()/of_node_put() unbalanced - destroy cset entry: attach overlay node /axi/csiss@80000000/ports/port@1/endpoint
[ 8191.912406] OF: ERROR: memory leak before free overlay changeset,  /axi/csiss@80000000/ports/port@0/endpoint
[ 8191.922328] OF: ERROR: memory leak, expected refcount 1 instead of -1073741824, of_node_get()/of_node_put() unbalanced - destroy cset entry: attach overlay node /axi/csiss@80000000/ports/port@0/endpoint
[ 8191.940542] OF: ERROR: memory leak, expected refcount 1 instead of 2, of_node_get()/of_node_put() unbalanced - destroy cset entry: attach overlay node /axi/i2c@80030000/i2c-mux@74/i2c@0/isp@3c/sensors/sensor@0
remove from slot 0 returns: 0 (Ok)

（三）啟動(dòng)Docker

先下載xilinx/smartcam的Docker IMG，通過docker pull xilinx/smartcam命令。

root@kria:~/Vitis-AI# docker pull xilinx/smartcam
Using default tag: latest
latest: Pulling from xilinx/smartcam
00f50047d606: Pull complete
d7951c234d55: Pull complete
05265a2d1f35: Pull complete
90b46a25b424: Pull complete
80e164c37cc5: Pull complete
3d8f42a1f194: Pull complete
b98fe3f03a5b: Pull complete
59a6d05de11d: Pull complete
c3201d2e9455: Pull complete
5a86aa1eda97: Pull complete
1c16e9132328: Pull complete
d5655ba163b7: Pull complete
3044adb41328: Pull complete
f5dc15e1f4ef: Pull complete
Digest: sha256:da2e52629011aeec332152a0f468d3ff156917dba9b596cf6d0de958d5dc29d7
Status: Downloaded newer image for xilinx/smartcam:latest
docker.io/xilinx/smartcam:latest

用了國(guó)內(nèi)的源，速度就是快，很快就下載完成。再看看imgs：

root@kria:~/Vitis-AI# docker images
REPOSITORY        TAG       IMAGE ID       CREATED         SIZE
hello-world       latest    b038788ddb22   4 months ago    9.14kB
xilinx/smartcam   latest    aa0270aef908   11 months ago   1.41GB

通過命令啟動(dòng)xilinx/smartcam鏡像，

此時(shí)root@xlnx-docker:/# 提示進(jìn)入Docker環(huán)境。

三、運(yùn)行

如文章之初所說，因?yàn)椴渴餠itis AI之后懶得poweroff插MIPI camera了（直接帶電插MIPI camera又感覺有風(fēng)險(xiǎn)），那就直接USB camera上吧。

先確定USB camera的名字，通過拔插USB攝像頭確定是/dev/media1，為什么不是/dev/video*？一個(gè)問號(hào)。

root@xlnx-docker:/# ls /dev/media*
/dev/media0  /dev/media1

本來想smartcam --usb=1 -W 640 -H 480 --target=dp，接上HDMI顯示器后不能顯示，貌似顯示器不支持這個(gè)分辨率，為什么是640X480這個(gè)奇葩的分辨率，是因?yàn)楹？礥SB攝像頭為720P，所以選了個(gè)攝像頭支持的分辨率。

root@xlnx-docker:/# smartcam --usb=1 -W 640 -H 480 --target=dp
Resize: mean_r=128.000000
Resize: mean_g=128.000000
Resize: mean_b=128.000000
Resize: scale_r=1.000000
Resize: scale_g=1.000000
Resize: scale_b=1.000000
[ 9484.157137] zynqmp-display fd4a0000.display: Layer width:height must be 640:480
[ 9484.187386] zynqmp-display fd4a0000.display: Layer width:height must be 640:480
[ 9484.202081] zynqmp-display fd4a0000.display: Layer width:height must be 640:480
[ 9484.224098] zynqmp-display fd4a0000.display: Layer width:height must be 640:480
[ 9484.264872] zynqmp-display fd4a0000.display: Layer width:height must be 640:480

那就RTSP吧，smartcam --usb=1 -W 640 -H 480 --target=rtsp伺候：

root@xlnx-docker:/# smartcam --usb=1 -W 640 -H 480 --target=rtsp
stream ready at:
 rtsp://172.17.0.1:554/test
rtsp://192.168.99.238:554/test

用支持RTSP的播放器再同一網(wǎng)段內(nèi)連接，本人頭像就不照了，剛好在刷抖音，刷個(gè)人臉來看看。

效果不錯(cuò)，最后，來一張合影,基本上各種線纜都接上了。