2022/02/20

NVIDIA L4T MLのDockerイメージで，Jetson上にCUDAをサポートするOpenCVの環境を準備する

概要

JetPack 4.6にはOpenCV 4.1.1が含まれていますが，CUDAサポートがOFFになっています． ここでは比較的手軽にCUDAサポートがONのOpenCVの環境を準備する方法として，NVIDIA L4T MLのDockerイメージを使用する方法をメモしておきます．

Jetson Jetson Nano
JetPack 4.6
NVIDIA L4T ML l4t-ml:r32.6.1-py3
NIVIDIA NGCとNVIDIA L4T ML

NVIDIA NGCでは，さまざまな用途のための，ディープラーニング，機械学習およびHPC向けのDockerイメージが配布されています． Jetson向けのDockerイメージは下記のリンクから一覧表示できます．

NVIDA NGCで配布されているDockerイメージの1つにNVIDIA L4T MLがあります．これはディープラーニング，機械学習やデータサイエンスの各種フレームワークをPython 3.6上で使用するための環境です． 下記ページの「Package Versions」を確認するとl4t-ml:r32.6.1-py3のリストにはOpenCV 4.5.0 (with CUDA)と記載されており，CUDAサポートがONになっていることがわかります．

Dockerコンテナを起動する
ワーキングディレクトリと起動スクリプトの準備

お好みの場所にワーキングディレクトリを作成し，その中にDockerコンテナの起動スクリプト，run.shを作成します． ここでは既存ディレクトリの/dataの配下にl4t-mlというディレクトリを作成しています．

1
2
3
$ cd /data
$ mkdir l4t-ml
$ cd l4t-ml/

お好みのテキストエディタでrun.shを作成し，下記リンクのrun.shの内容をコピー・アンド・ペーストしてください． なお，使用するDockerイメージのタグはJetPack 4.6に対応するr32.6.1-py3になっています．違うタグを指定したい場合は，NVIDIA L4T MLのページを参照して任意のタグに置き換えてください．

1
$ vi run.sh
Dockerコンテナの起動

少なくともこれ以降の作業は，Jetsonにキーボード，マウスおよびディスプレイを接続して直接GUIログインしていることを前提に説明します． run.shに実行権限を付与し，実行してください．初回はDockerイメージのPullがあり，Dockerコンテナの起動に時間がかかります．

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
$ chmod a+x ./run.sh
$ ./run.sh
access control disabled, clients can connect from any host
[sudo] password for username:
Unable to find image 'nvcr.io/nvidia/l4t-ml:r32.6.1-py3' locally
r32.6.1-py3: Pulling from nvidia/l4t-ml
e47a8a86d66c: Pulling fs layer
bdce3430dad6: Pulling fs layer
c26a6b81c746: Pulling fs layer
a70635e646de: Waiting
24cbe60e3161: Waiting
c7f64cc97a39: Waiting
8777adb92eda: Waiting
542a24b3572f: Waiting
3d5ade2b8849: Waiting
3e865584f789: Waiting
4811af6cacf1: Waiting
db645757aac7: Pull complete
0105dc23dc93: Pull complete
c8aff3baf097: Pull complete
5f50b3e38480: Pull complete
b5418d85b074: Pull complete
d09ffaf0f11a: Pull complete
f1cf5625978f: Pull complete
514fd4c3f8c5: Pull complete
0988d0ccf8a6: Pull complete
ffce97733d6c: Pull complete
80d5eedfe895: Pull complete
d19792019860: Pull complete
3ee7b31a8890: Pull complete
8aca18eb38a5: Pull complete
153000266b1a: Pull complete
23bf7741f5d8: Pull complete
d03fa2111a1c: Pull complete
095c23b57578: Pull complete
2de9347ccaf2: Pull complete
0f26c4b0a765: Pull complete
ee3bc588c8ee: Pull complete
aa25eb953d98: Pull complete
97fd80a69cc6: Pull complete
Digest: sha256:1f4ef02343223cab6bbe5bf85d5575364b7ee0ee23d72c2e3b89416ca3806f7e
Status: Downloaded newer image for nvcr.io/nvidia/l4t-ml:r32.6.1-py3
allow 10 sec for JupyterLab to start @ http://192.168.110.36:8888 (password nvidia)
JupterLab logging location:  /var/log/jupyter.log  (inside the container)
root@jetson-nano:/#

Dockerコンテナが起動し，Dockerコンテナの端末が表示されました．

dockerコマンドを実行するとエラーになる場合は，下記が解決の参考になるかもしれません．

Python 3でOpenCVライブラリの情報を確認する

下記の投稿で紹介した方法で，Dockerコンテナ上のOpenCVライブラリの情報を確認してみました． CUDAサポートがONになっています．

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
root@jetson-nano:/# python3
Python 3.6.9 (default, Jan 26 2021, 15:33:00)
[GCC 8.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import cv2
>>> print(cv2.__version__)
4.5.0
>>> print(cv2.getBuildInformation())
 
General configuration for OpenCV 4.5.0 =====================================
  Version control:               4.5.0
 
  Extra modules:
    Location (extra):            /opt/opencv_contrib/modules
    Version control (extra):     4.5.0
 
  Platform:
    Timestamp:                   2021-07-19T21:05:54Z
    Host:                        Linux 4.9.201-tegra aarch64
    CMake:                       3.10.2
    CMake generator:             Unix Makefiles
    CMake build tool:            /usr/bin/make
    Configuration:               RELEASE
 
  CPU/HW features:
    Baseline:                    NEON FP16
      required:                  NEON
      disabled:                  VFPV3
 
  C/C++:
    Built as dynamic libs?:      YES
    C++ standard:                11
    C++ Compiler:                /usr/bin/c++  (ver 7.5.0)
    C++ flags (Release):         -fsigned-char -W -Wall -Werror=return-type -Werror=non-virtual-dtor -Werror=address -Werror=sequence-point -Wformat -Werror=format-security -Wmissing-declarations -Wundef -Winit-self -Wpointer-arith -Wshadow -Wsign-promo -Wuninitialized -Winit-self -Wsuggest-override -Wno-delete-non-virtual-dtor -Wno-comment -Wimplicit-fallthrough=3 -Wno-strict-overflow -fdiagnostics-show-option -pthread -fomit-frame-pointer -ffunction-sections -fdata-sections    -fvisibility=hidden -fvisibility-inlines-hidden -O3 -DNDEBUG  -DNDEBUG
    C++ flags (Debug):           -fsigned-char -W -Wall -Werror=return-type -Werror=non-virtual-dtor -Werror=address -Werror=sequence-point -Wformat -Werror=format-security -Wmissing-declarations -Wundef -Winit-self -Wpointer-arith -Wshadow -Wsign-promo -Wuninitialized -Winit-self -Wsuggest-override -Wno-delete-non-virtual-dtor -Wno-comment -Wimplicit-fallthrough=3 -Wno-strict-overflow -fdiagnostics-show-option -pthread -fomit-frame-pointer -ffunction-sections -fdata-sections    -fvisibility=hidden -fvisibility-inlines-hidden -g  -O0 -DDEBUG -D_DEBUG
    C Compiler:                  /usr/bin/cc
    C flags (Release):           -fsigned-char -W -Wall -Werror=return-type -Werror=address -Werror=sequence-point -Wformat -Werror=format-security -Wmissing-declarations -Wmissing-prototypes -Wstrict-prototypes -Wundef -Winit-self -Wpointer-arith -Wshadow -Wuninitialized -Winit-self -Wno-comment -Wimplicit-fallthrough=3 -Wno-strict-overflow -fdiagnostics-show-option -pthread -fomit-frame-pointer -ffunction-sections -fdata-sections    -fvisibility=hidden -O3 -DNDEBUG  -DNDEBUG
    C flags (Debug):             -fsigned-char -W -Wall -Werror=return-type -Werror=address -Werror=sequence-point -Wformat -Werror=format-security -Wmissing-declarations -Wmissing-prototypes -Wstrict-prototypes -Wundef -Winit-self -Wpointer-arith -Wshadow -Wuninitialized -Winit-self -Wno-comment -Wimplicit-fallthrough=3 -Wno-strict-overflow -fdiagnostics-show-option -pthread -fomit-frame-pointer -ffunction-sections -fdata-sections    -fvisibility=hidden -g  -O0 -DDEBUG -D_DEBUG
    Linker flags (Release):      -Wl,--gc-sections -Wl,--as-needed 
    Linker flags (Debug):        -Wl,--gc-sections -Wl,--as-needed 
    ccache:                      NO
    Precompiled headers:         NO
    Extra dependencies:          m pthread cudart_static dl rt nppc nppial nppicc nppicom nppidei nppif nppig nppim nppist nppisu nppitc npps cublas cudnn cufft -L/usr/local/cuda/lib64 -L/usr/lib/aarch64-linux-gnu
    3rdparty dependencies:
 
  OpenCV modules:
    To be built:                 alphamat aruco bgsegm bioinspired calib3d ccalib core cudaarithm cudabgsegm cudacodec cudafeatures2d cudafilters cudaimgproc cudalegacy cudaobjdetect cudaoptflow cudastereo cudawarping cudev datasets dnn dnn_objdetect dnn_superres dpm face features2d flann freetype fuzzy gapi hfs highgui img_hash imgcodecs imgproc intensity_transform line_descriptor mcc ml objdetect optflow phase_unwrapping photo plot python3 quality rapid reg rgbd saliency shape stereo stitching structured_light superres surface_matching text tracking video videoio videostab xfeatures2d ximgproc xobjdetect xphoto
    Disabled:                    world
    Disabled by dependency:      -
    Unavailable:                 cnn_3dobj cvv hdf java js julia matlab ovis python2 sfm ts viz
    Applications:                apps
    Documentation:               NO
    Non-free algorithms:         YES
 
  GUI:
    GTK+:                        YES (ver 3.22.30)
      GThread :                  YES (ver 2.56.4)
      GtkGlExt:                  NO
    OpenGL support:              NO
    VTK support:                 NO
 
  Media I/O:
    ZLib:                        /usr/lib/aarch64-linux-gnu/libz.so (ver 1.2.11)
    JPEG:                        /usr/lib/aarch64-linux-gnu/libjpeg.so (ver 80)
    WEBP:                        build (ver encoder: 0x020f)
    PNG:                         /usr/lib/aarch64-linux-gnu/libpng.so (ver 1.6.34)
    TIFF:                        build (ver 42 - 4.0.10)
    JPEG 2000:                   build (ver 2.3.1)
    OpenEXR:                     build (ver 2.3.0)
    HDR:                         YES
    SUNRASTER:                   YES
    PXM:                         YES
    PFM:                         YES
 
  Video I/O:
    DC1394:                      YES (2.2.5)
    FFMPEG:                      YES
      avcodec:                   YES (57.107.100)
      avformat:                  YES (57.83.100)
      avutil:                    YES (55.78.100)
      swscale:                   YES (4.8.100)
      avresample:                YES (3.7.0)
    GStreamer:                   YES (1.14.5)
    v4l/v4l2:                    YES (linux/videodev2.h)
 
  Parallel framework:            TBB (ver 2017.0 interface 9107)
 
  Trace:                         YES (with Intel ITT)
 
  Other third-party libraries:
    Lapack:                      YES (/usr/lib/aarch64-linux-gnu/liblapack.so /usr/lib/aarch64-linux-gnu/libcblas.so /usr/lib/aarch64-linux-gnu/libatlas.so)
    Eigen:                       YES (ver 3.3.4)
    Custom HAL:                  YES (carotene (ver 0.0.1))
    Protobuf:                    build (3.5.1)
 
  NVIDIA CUDA:                   YES (ver 10.2, CUFFT CUBLAS FAST_MATH)
    NVIDIA GPU arch:             53 62 72
    NVIDIA PTX archs:
 
  cuDNN:                         YES (ver 8.0.0)
 
  Python 3:
    Interpreter:                 /usr/bin/python3 (ver 3.6.9)
    Libraries:                   /usr/lib/aarch64-linux-gnu/libpython3.6m.so (ver 3.6.9)
    numpy:                       /usr/lib/python3/dist-packages/numpy/core/include (ver 1.13.3)
    install path:                lib/python3.6/dist-packages/cv2/python-3.6
 
  Python (for build):            /usr/bin/python2.7
 
  Java:                         
    ant:                         NO
    JNI:                         NO
    Java wrappers:               NO
    Java tests:                  NO
 
  Install to:                    /usr/local
-----------------------------------------------------------------
 
 
>>> exit()

製品紹介

Labels: ,

0 件のコメント:

コメントを投稿