Getting Started with Nvidia Jetson Nano


Series

Guide

Jetson Family

  • Jetson TX1 Developer Kit
  • Jetson TX2 Developer Kit
  • Jetson AGX Xaiver Developer Kit
  • Jetson Nano Developer Kit

SDKs and Tools

  • NVIDIA JetPack
  • NVIDIA DeepStream SDK
  • NVIDIA DIGITS for training

JetPack includes:
• Full desktop Linux with NVIDIA drivers
• AI and Computer Vision libraries and APIs
• Developer tools
• Documentation and sample code

Training GPU:

  • Maxwell, Pascal, Volta, or Turing-based GPU (ideally with at least 6GB video memory) optionally, AWS P2/P3 instance or Microsoft Azure N-series
  • Ubuntu 16.04/18.04 x86_64

Deployment:

  • Jetson Nano Developer Kit with JetPack 4.2 or newer (Ubuntu 18.04 aarch64).
  • Jetson Xavier Developer Kit with JetPack 4.0 or newer (Ubuntu 18.04 aarch64)
  • Jetson TX2 Developer Kit with JetPack 3.0 or newer (Ubuntu 16.04 aarch64).
  • Jetson TX1 Developer Kit with JetPack 2.3 or newer (Ubuntu 16.04 aarch64).

Jetson Nano Developer Kit

intro

Jetson Nano was introduced in April 2019 for only $99.

jetson nano
Jetson Nano image

  1. microSD card slot for main storage
  2. 40-pin expansion header
  3. Micro-USB port for 5V power input or for data
  4. Gigabit Ethernet port
  5. USB 3.0 ports (x4)
  6. HDMI output port
  7. DisplayPort connector
  8. DC Barrel jack for 5V power input
  9. MIPI CSI camera connector

power input: 3 and 8
camera: 9 (MIPI CSI camera)
green LED (D53) close to the micro USB port should turn green

inference performance
Jetson Nano inference performance

multiple cameras with jetson nano

Write Image to the microSD Card

  1. Download the jetson-nano-sd-card-image-r3223.zip
  2. Format the microSD card to ExFAT if it’s a 64Gb or higher card, and to FAT if it’s less.
  3. Use etcher or linux command to write image to microSD.

    Image size about 5G, so be patient to download

linux command to write image to microSD

$ df -h 
Filesystem      Size  Used Avail Use% Mounted on
udev            7.8G     0  7.8G   0% /dev
tmpfs           1.6G   18M  1.6G   2% /run
/dev/sdb6       184G  162G   13G  93% /
tmpfs           7.8G   71M  7.8G   1% /dev/shm
tmpfs           5.0M  4.0K  5.0M   1% /run/lock
tmpfs           7.8G     0  7.8G   0% /sys/fs/cgroup
/dev/sdb5       453M  157M  270M  37% /boot
tmpfs           1.6G   56K  1.6G   1% /run/user/1000
/dev/sdb4       388G  337G   52G  87% /media/kezunlin/Workspace
/dev/sdc1        30G   32K   30G   1% /media/kezunlin/nano


$ dmesg | tail | awk '$3 == "sd" {print}'

#In this example, we can see the 32GB microSD card was assigned /dev/sdc:
[  613.537818] sd 4:0:0:0: Attached scsi generic sg2 type 0
[  613.940079] sd 4:0:0:0: [sdc] 62333952 512-byte logical blocks: (31.9 GB/29.7 GiB)
[  613.940664] sd 4:0:0:0: [sdc] Write Protect is off
[  613.940672] sd 4:0:0:0: [sdc] Mode Sense: 87 00 00 00
[  613.942730] sd 4:0:0:0: [sdc] Write cache: disabled, read cache: enabled, doesnt support DPO or FUA
[  613.956666] sd 4:0:0:0: [sdc] Attached SCSI removable disk

# Use this command to write the zipped SD card image to the microSD card:

$ /usr/bin/unzip -p ~/Downloads/jetson-nano-sd-card-image-r3223.zip | sudo /bin/dd of=/dev/sdc bs=1M status=progress

0+167548 records in
0+167548 records out
12884901888 bytes (13 GB, 12 GiB) copied, 511.602 s, 25.2 MB/s


# 12 partitions generated by the writing process ???
$ sudo fdisk -l 

GPT PMBR size mismatch (25165823 != 62333951) will be corrected by w(rite).
Disk /dev/sde: 29.7 GiB, 31914983424 bytes, 62333952 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: gpt
Disk identifier: E696E264-F2EA-434A-900C-D9ACA2F99E43

Device     Start      End  Sectors  Size Type
/dev/sde1  24576 25165790 25141215   12G Linux filesystem
/dev/sde2   2048     2303      256  128K Linux filesystem
/dev/sde3   4096     4991      896  448K Linux filesystem
/dev/sde4   6144     7295     1152  576K Linux filesystem
/dev/sde5   8192     8319      128   64K Linux filesystem
/dev/sde6  10240    10623      384  192K Linux filesystem
/dev/sde7  12288    13439     1152  576K Linux filesystem
/dev/sde8  14336    14463      128   64K Linux filesystem
/dev/sde9  16384    17663     1280  640K Linux filesystem
/dev/sde10 18432    19327      896  448K Linux filesystem
/dev/sde11 20480    20735      256  128K Linux filesystem
/dev/sde12 22528    22687      160   80K Linux filesystem

Partition table entries are not in disk order.


# When the dd command finishes, eject the disk device from the command line:
$ sudo eject /dev/sdc

# Physically remove microSD card from the computer.

Steps:

  1. Insert the microSD card into the appropriate slot
  2. Connect the display and USB keyboard /mouse and Ethernet cable.
  3. Depending on the power supply you want to use, you may have to add or remove the jumper for power selection:
    – If using a jack(part 8), the jumper must be set.
    – if using USB (part 3), the jumper must be off.
  4. Plug in the power supply. The green LED (D53) close to the micro USB port should turn green, and the display should show the NVIDIA logo before booting begins.

Prepare Nano System

  • Jetson Nano L4T 32.2.1-20190812212815 (JetPack 4.2.2)
  • Ubuntu 18.04 aarch64 (bionic)
  • CUDA 10.0 SM_72 (installed)
  • TensorRT-5.1.6.0 (installed)
  • OpenCV 3.3.1 (installed)
  • Python 2.7 (installed)
  • Python 3.6.9
  • Numpy 1.13.3
  • QT 5.9.5

CUDA-10.0 and TensorRT-5.1.6.0 has already installed on Jetson Nano.
ARM 64 machines such as the Jetson Nano and Raspberry Pi.

SSH for nano with ForwardX11

for nano, edit /etc/ssh/ssh_config

X11Forwarding yes

and restart ssh

$ sudo /etc/init.d/ssh restart
[sudo] password for nano: 
[ ok ] Restarting ssh (via systemctl): ssh.service.

for my ubuntu client

edit ~/.ssh/config

Host nano
  HostName 192.168.0.63
  User nano
  ForwardX11 yes

ForwardX11 yes

# method 1: `~/.ssh/config` with  `ForwardX11 yes`
$ ssh [email protected]

# method 2: with `-X`
$ ssh -X [email protected] 
# `-X` means enabling ForwardX11

check versions

$ uname -a 
Linux nano-desktop 4.9.140-tegra #1 SMP PREEMPT Sat Oct 19 15:54:06 PDT 2019 aarch64 aarch64 aarch64 GNU/Linux

$ dpkg-query --show nvidia-l4t-core
nvidia-l4t-core    32.2.1-20190812212815

$ python --version
Python 2.7.15+

$ git --version
git version 2.17.1

# check tensorrt version
$ ll -al /usr/lib/aarch64-linux-gnu/libnvinfer_plugin.so.5
lrwxrwxrwx 1 root root 26 Jun  5  2019 /usr/lib/aarch64-linux-gnu/libnvinfer_plugin.so.5 -> libnvinfer_plugin.so.5.1.6


# edit `~.bashrc`
# Add this to your .bashrc file

export CUDA_HOME=/usr/local/cuda
# Adds the CUDA compiler to the PATH
export PATH=$CUDA_HOME/bin:$PATH
# Adds the libraries
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH

# check cuda version
$ source ~/.bashrc
$ nvcc --version

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Mon_Mar_11_22:13:24_CDT_2019
Cuda compilation tools, release 10.0, V10.0.326

NO NEED TO download TensorRT 5.1.5.0 GA for Ubuntu 18.04 and CUDA 10.0 tar package from here and place at /opt/TensorRT-5.1.5.0

install packages

  • cmake 3.10.2
  • cmake-gui 3.10.2
  • python 3.6.9
  • QT 5.9.5

use official ubuntu source
do not replace /etc/opt/sources.list with aliyun source,otherwise lots of packages will fail to install

# update
$ sudo apt-get update

# install cmake
$ sudo apt-get install cmake

Get:1 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 cmake arm64 3.10.2-1ubuntu2.18.04.1 [2,971 kB]

# install cmake-gui 
$ sudo apt-get install cmake-gui cmake-qt-gui

Get:1 http://ports.ubuntu.com/ubuntu-ports bionic-updates/universe arm64 cmake-qt-gui arm64 3.10.2-1ubuntu2.18.04.1 [1,527 kB]


# install python 3.6.9
$ sudo apt -y install libpython3-dev python3-numpy

Get:1 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 libpython3.6 arm64 3.6.9-1~18.04 [1,307 kB]
Get:2 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 python3.6 arm64 3.6.9-1~18.04 [203 kB]                                           
Get:3 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 libpython3.6-stdlib arm64 3.6.9-1~18.04 [1,609 kB]                               
Get:4 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 python3.6-minimal arm64 3.6.9-1~18.04 [1,327 kB]                                 
Get:5 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 libpython3.6-minimal arm64 3.6.9-1~18.04 [528 kB]                                
Get:6 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 libpython3.6-dev arm64 3.6.9-1~18.04 [45.0 MB]                                   
Get:7 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 libpython3-dev arm64 3.6.7-1~18.04 [7,328 B]                                     
Get:8 http://ports.ubuntu.com/ubuntu-ports bionic/main arm64 python3-numpy arm64 1:1.13.3-2ubuntu1 [1,734 kB]

# install qt5
$ sudo apt-get install qtbase5-dev

Get:1 http://ports.ubuntu.com/ubuntu-ports bionic-updates/main arm64 libqt5core5a arm64 5.9.5+dfsg-0ubuntu2.4 [1933 kB]

packages conclusion

sudo apt -y install cmake cmake-gui cmake-qt-gui
sudo apt -y install libpython3-dev python3-numpy python3-pip
sudo apt -y install qtbase5-dev

# other packages for c++ programs
sudo apt -y install libcrypto++-dev 
sudo apt -y install libgoogle-log-dev libgflags-dev
sudo apt -y install --no-install-recommends libboost-all-dev

Build jetson-inference

see jetson-inference

cmake and configure

git clone --recursive https://github.com/dusty-nv/jetson-inference
cd jetson-inference
git submodule update --init
mkdir build
cd build
sudo cmake ..

congigure

do not donwload models
do not download pytorch

run download models and pytorch later if you really need

  $ cd jetson-inference/tools
  $ ./download-models.sh

  $ cd jetson-inference/build
  $ ./install-pytorch.sh

or download models from here
wget -b -c https://github.com/dusty-nv/jetson-inference/releases/download/model-mirror-190618/ResNet-18.tar.gz

congigure output
cmake-gui jetson-inference

[jetson-inference]  Checking for 'dialog' deb package...installed
[jetson-inference]  FOUND_DIALOG=INSTALLED

[jetson-inference]  Model selection status:  0
[jetson-inference]  No models were selected for download.

[jetson-inference]  to run this tool again, use the following commands:
    $ cd <jetson-inference>/tools
    $ ./download-models.sh

[jetson-inference]  Checking for 'dialog' deb package...installed
[jetson-inference]  FOUND_DIALOG=INSTALLED

head: cannot open '/etc/nv_tegra_release' for reading: No such file or directory
[jetson-inference]  reading L4T version from "dpkg-query --show nvidia-l4t-core"
[jetson-inference]  Jetson BSP Version:  L4T R32.2

[jetson-inference]  Package selection status:  1
[jetson-inference]  Package selection cancelled.

[jetson-inference]  installation complete, exiting with status code 0
[jetson-inference]  to run this tool again, use the following commands:
    $ cd <jetson-inference>/build
    $ ./install-pytorch.sh

[Pre-build]  Finished CMakePreBuild script

-- Finished installing dependencies
-- using patched FindCUDA.cmake
Looking for pthread.h
Looking for pthread.h - found
Looking for pthread_create
Looking for pthread_create - not found
Looking for pthread_create in pthreads
Looking for pthread_create in pthreads - not found
Looking for pthread_create in pthread
Looking for pthread_create in pthread - found
Found Threads: TRUE  

-- using patched FindCUDA.cmake
-- CUDA version: 10.0
-- CUDA 10 detected, enabling SM_72
-- OpenCV version:  3.3.1
-- OpenCV version >= 3.0.0, enabling OpenCV
-- system arch:  aarch64
-- output path:  /home/nano/git/jetson-inference/build/aarch64
-- Copying /home/nano/git/jetson-inference/c/detectNet.h
-- Copying /home/nano/git/jetson-inference/c/homographyNet.h
-- Copying /home/nano/git/jetson-inference/c/imageNet.h
-- Copying /home/nano/git/jetson-inference/c/segNet.h
-- Copying /home/nano/git/jetson-inference/c/superResNet.h
-- Copying /home/nano/git/jetson-inference/c/tensorNet.h
-- Copying /home/nano/git/jetson-inference/c/imageNet.cuh
-- Copying /home/nano/git/jetson-inference/calibration/randInt8Calibrator.h
Could NOT find Doxygen (missing: DOXYGEN_EXECUTABLE) 
-- found Qt5Widgets version:  5.9.5
-- found Qt5Widgets defines:  -DQT_WIDGETS_LIB;-DQT_GUI_LIB;-DQT_CORE_LIB
-- found Qt5Widgets library:  Qt5::Widgets
-- found Qt5Widgets include:  /usr/include/aarch64-linux-gnu/qt5/;/usr/include/aarch64-linux-gnu/qt5/QtWidgets;/usr/include/aarch64-linux-gnu/qt5/QtGui;/usr/include/aarch64-linux-gnu/qt5/QtCore;/usr/lib/aarch64-linux-gnu/qt5//mkspecs/linux-g++
-- camera-capture:  building as submodule, /home/nano/git/jetson-inference/tools
-- jetson-utils:  building as submodule, /home/nano/git/jetson-inference
-- Copying /home/nano/git/jetson-inference/utils/XML.h
-- Copying /home/nano/git/jetson-inference/utils/commandLine.h
-- Copying /home/nano/git/jetson-inference/utils/filesystem.h
-- Copying /home/nano/git/jetson-inference/utils/mat33.h
-- Copying /home/nano/git/jetson-inference/utils/pi.h
-- Copying /home/nano/git/jetson-inference/utils/rand.h
-- Copying /home/nano/git/jetson-inference/utils/timespec.h
-- Copying /home/nano/git/jetson-inference/utils/camera/gstCamera.h
-- Copying /home/nano/git/jetson-inference/utils/camera/v4l2Camera.h
-- Copying /home/nano/git/jetson-inference/utils/codec/gstDecoder.h
-- Copying /home/nano/git/jetson-inference/utils/codec/gstEncoder.h
-- Copying /home/nano/git/jetson-inference/utils/codec/gstUtility.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaFont.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaMappedMemory.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaNormalize.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaOverlay.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaRGB.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaResize.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaUtility.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaWarp.h
-- Copying /home/nano/git/jetson-inference/utils/cuda/cudaYUV.h
-- Copying /home/nano/git/jetson-inference/utils/display/glDisplay.h
-- Copying /home/nano/git/jetson-inference/utils/display/glTexture.h
-- Copying /home/nano/git/jetson-inference/utils/display/glUtility.h
-- Copying /home/nano/git/jetson-inference/utils/image/imageIO.h
-- Copying /home/nano/git/jetson-inference/utils/image/loadImage.h
-- Copying /home/nano/git/jetson-inference/utils/input/devInput.h
-- Copying /home/nano/git/jetson-inference/utils/input/devJoystick.h
-- Copying /home/nano/git/jetson-inference/utils/input/devKeyboard.h
-- Copying /home/nano/git/jetson-inference/utils/network/Endian.h
-- Copying /home/nano/git/jetson-inference/utils/network/IPv4.h
-- Copying /home/nano/git/jetson-inference/utils/network/NetworkAdapter.h
-- Copying /home/nano/git/jetson-inference/utils/network/Socket.h
-- Copying /home/nano/git/jetson-inference/utils/threads/Event.h
-- Copying /home/nano/git/jetson-inference/utils/threads/Mutex.h
-- Copying /home/nano/git/jetson-inference/utils/threads/Process.h
-- Copying /home/nano/git/jetson-inference/utils/threads/Thread.h
-- trying to build Python bindings for Python versions:  2.7;3.6;3.7
-- detecting Python 2.7...
-- found Python version:  2.7 (2.7.15+)
-- found Python include:  /usr/include/python2.7
-- found Python library:  /usr/lib/aarch64-linux-gnu/libpython2.7.so
-- CMake module path:  /home/nano/git/jetson-inference/utils/cuda;/home/nano/git/jetson-inference/utils/python/bindings
NumPy ver. 1.13.3 found (include: /usr/lib/python2.7/dist-packages/numpy/core/include)
-- found NumPy version:  1.13.3
-- found NumPy include:  /usr/lib/python2.7/dist-packages/numpy/core/include
-- detecting Python 3.6...
-- found Python version:  3.6 (3.6.9)
-- found Python include:  /usr/include/python3.6m
-- found Python library:  /usr/lib/aarch64-linux-gnu/libpython3.6m.so
-- CMake module path:  /home/nano/git/jetson-inference/utils/cuda;/home/nano/git/jetson-inference/utils/python/bindings
NumPy ver. 1.13.3 found (include: /usr/lib/python3/dist-packages/numpy/core/include)
-- found NumPy version:  1.13.3
-- found NumPy include:  /usr/lib/python3/dist-packages/numpy/core/include
-- detecting Python 3.7...
-- Python 3.7 wasn't found
-- Copying /home/nano/git/jetson-inference/utils/python/examples/camera-viewer.py
-- Copying /home/nano/git/jetson-inference/utils/python/examples/cuda-from-numpy.py
-- Copying /home/nano/git/jetson-inference/utils/python/examples/cuda-to-numpy.py
-- Copying /home/nano/git/jetson-inference/utils/python/examples/gl-display-test.py
-- trying to build Python bindings for Python versions:  2.7;3.6;3.7
-- detecting Python 2.7...
-- found Python version:  2.7 (2.7.15+)
-- found Python include:  /usr/include/python2.7
-- found Python library:  /usr/lib/aarch64-linux-gnu/libpython2.7.so
-- detecting Python 3.6...
-- found Python version:  3.6 (3.6.9)
-- found Python include:  /usr/include/python3.6m
-- found Python library:  /usr/lib/aarch64-linux-gnu/libpython3.6m.so
-- detecting Python 3.7...
-- Python 3.7 wasn't found
-- Copying /home/nano/git/jetson-inference/python/examples/detectnet-camera.py
-- Copying /home/nano/git/jetson-inference/python/examples/detectnet-console.py
-- Copying /home/nano/git/jetson-inference/python/examples/imagenet-camera.py
-- Copying /home/nano/git/jetson-inference/python/examples/imagenet-console.py
-- Copying /home/nano/git/jetson-inference/python/examples/my-detection.py
-- Copying /home/nano/git/jetson-inference/python/examples/my-recognition.py
-- Copying /home/nano/git/jetson-inference/python/examples/segnet-batch.py
-- Copying /home/nano/git/jetson-inference/python/examples/segnet-camera.py
-- Copying /home/nano/git/jetson-inference/python/examples/segnet-console.py
Configuring done

compile and install

generate and compile

sudo make

install jetson-inference

sudo make install
sudo ldconfig

output

[  1%] Linking CXX shared library ../aarch64/lib/libjetson-utils.so
[ 31%] Built target jetson-utils
[ 32%] Linking CXX shared library aarch64/lib/libjetson-inference.so
[ 43%] Built target jetson-inference
[ 44%] Linking CXX executable ../../aarch64/bin/imagenet-console
[ 45%] Built target imagenet-console
[ 46%] Linking CXX executable ../../aarch64/bin/imagenet-camera
[ 47%] Built target imagenet-camera
[ 47%] Linking CXX executable ../../aarch64/bin/detectnet-console
[ 48%] Built target detectnet-console
[ 49%] Linking CXX executable ../../aarch64/bin/detectnet-camera
[ 50%] Built target detectnet-camera
[ 50%] Linking CXX executable ../../aarch64/bin/segnet-console
[ 51%] Built target segnet-console
[ 52%] Linking CXX executable ../../aarch64/bin/segnet-camera
[ 53%] Built target segnet-camera
[ 54%] Linking CXX executable ../../aarch64/bin/superres-console
[ 55%] Built target superres-console
[ 56%] Linking CXX executable ../../aarch64/bin/homography-console
[ 57%] Built target homography-console
[ 58%] Linking CXX executable ../../aarch64/bin/homography-camera
[ 59%] Built target homography-camera
[ 60%] Automatic MOC for target camera-capture
[ 60%] Built target camera-capture_autogen
[ 61%] Linking CXX executable ../../aarch64/bin/camera-capture
[ 64%] Built target camera-capture
[ 65%] Linking CXX executable ../../aarch64/bin/trt-bench
[ 66%] Built target trt-bench
[ 67%] Linking CXX executable ../../aarch64/bin/trt-console
[ 68%] Built target trt-console
[ 69%] Linking CXX executable ../../../aarch64/bin/camera-viewer
[ 70%] Built target camera-viewer
[ 71%] Linking CXX executable ../../../aarch64/bin/v4l2-console
[ 72%] Built target v4l2-console
[ 73%] Linking CXX executable ../../../aarch64/bin/v4l2-display
[ 74%] Built target v4l2-display
[ 75%] Linking CXX executable ../../../aarch64/bin/gl-display-test
[ 76%] Built target gl-display-test
[ 76%] Linking CXX shared library ../../../aarch64/lib/python/2.7/jetson_utils_python.so
[ 82%] Built target jetson-utils-python-27
[ 83%] Linking CXX shared library ../../../aarch64/lib/python/3.6/jetson_utils_python.so
[ 89%] Built target jetson-utils-python-36
[ 90%] Linking CXX shared library ../../aarch64/lib/python/2.7/jetson_inference_python.so
[ 95%] Built target jetson-inference-python-27
[ 96%] Linking CXX shared library ../../aarch64/lib/python/3.6/jetson_inference_python.so
[100%] Built target jetson-inference-python-36
Install the project...
-- Install configuration: ""
-- Installing: /usr/local/include/jetson-inference/detectNet.h
-- Installing: /usr/local/include/jetson-inference/homographyNet.h
-- Installing: /usr/local/include/jetson-inference/imageNet.h
-- Installing: /usr/local/include/jetson-inference/segNet.h
-- Installing: /usr/local/include/jetson-inference/superResNet.h
-- Installing: /usr/local/include/jetson-inference/tensorNet.h
-- Installing: /usr/local/include/jetson-inference/imageNet.cuh
-- Installing: /usr/local/include/jetson-inference/randInt8Calibrator.h
-- Installing: /usr/local/lib/libjetson-inference.so
-- Set runtime path of "/usr/local/lib/libjetson-inference.so" to ""
-- Installing: /usr/local/share/jetson-inference/cmake/jetson-inferenceConfig.cmake
-- Installing: /usr/local/share/jetson-inference/cmake/jetson-inferenceConfig-noconfig.cmake
-- Installing: /usr/local/bin/imagenet-console
-- Set runtime path of "/usr/local/bin/imagenet-console" to ""
-- Installing: /usr/local/bin/imagenet-camera
-- Set runtime path of "/usr/local/bin/imagenet-camera" to ""
-- Installing: /usr/local/bin/detectnet-console
-- Set runtime path of "/usr/local/bin/detectnet-console" to ""
-- Installing: /usr/local/bin/detectnet-camera
-- Set runtime path of "/usr/local/bin/detectnet-camera" to ""
-- Installing: /usr/local/bin/segnet-console
-- Set runtime path of "/usr/local/bin/segnet-console" to ""
-- Installing: /usr/local/bin/segnet-camera
-- Set runtime path of "/usr/local/bin/segnet-camera" to ""
-- Installing: /usr/local/bin/superres-console
-- Set runtime path of "/usr/local/bin/superres-console" to ""
-- Installing: /usr/local/bin/homography-console
-- Set runtime path of "/usr/local/bin/homography-console" to ""
-- Installing: /usr/local/bin/homography-camera
-- Set runtime path of "/usr/local/bin/homography-camera" to ""
-- Installing: /usr/local/bin/camera-capture
-- Set runtime path of "/usr/local/bin/camera-capture" to ""
-- Installing: /usr/local/include/jetson-utils/XML.h
-- Installing: /usr/local/include/jetson-utils/commandLine.h
-- Installing: /usr/local/include/jetson-utils/filesystem.h
-- Installing: /usr/local/include/jetson-utils/mat33.h
-- Installing: /usr/local/include/jetson-utils/pi.h
-- Installing: /usr/local/include/jetson-utils/rand.h
-- Installing: /usr/local/include/jetson-utils/timespec.h
-- Installing: /usr/local/include/jetson-utils/gstCamera.h
-- Installing: /usr/local/include/jetson-utils/v4l2Camera.h
-- Installing: /usr/local/include/jetson-utils/gstDecoder.h
-- Installing: /usr/local/include/jetson-utils/gstEncoder.h
-- Installing: /usr/local/include/jetson-utils/gstUtility.h
-- Installing: /usr/local/include/jetson-utils/cudaFont.h
-- Installing: /usr/local/include/jetson-utils/cudaMappedMemory.h
-- Installing: /usr/local/include/jetson-utils/cudaNormalize.h
-- Installing: /usr/local/include/jetson-utils/cudaOverlay.h
-- Installing: /usr/local/include/jetson-utils/cudaRGB.h
-- Installing: /usr/local/include/jetson-utils/cudaResize.h
-- Installing: /usr/local/include/jetson-utils/cudaUtility.h
-- Installing: /usr/local/include/jetson-utils/cudaWarp.h
-- Installing: /usr/local/include/jetson-utils/cudaYUV.h
-- Installing: /usr/local/include/jetson-utils/glDisplay.h
-- Installing: /usr/local/include/jetson-utils/glTexture.h
-- Installing: /usr/local/include/jetson-utils/glUtility.h
-- Installing: /usr/local/include/jetson-utils/imageIO.h
-- Installing: /usr/local/include/jetson-utils/loadImage.h
-- Installing: /usr/local/include/jetson-utils/devInput.h
-- Installing: /usr/local/include/jetson-utils/devJoystick.h
-- Installing: /usr/local/include/jetson-utils/devKeyboard.h
-- Installing: /usr/local/include/jetson-utils/Endian.h
-- Installing: /usr/local/include/jetson-utils/IPv4.h
-- Installing: /usr/local/include/jetson-utils/NetworkAdapter.h
-- Installing: /usr/local/include/jetson-utils/Socket.h
-- Installing: /usr/local/include/jetson-utils/Event.h
-- Installing: /usr/local/include/jetson-utils/Mutex.h
-- Installing: /usr/local/include/jetson-utils/Process.h
-- Installing: /usr/local/include/jetson-utils/Thread.h
-- Installing: /usr/local/lib/libjetson-utils.so
-- Installing: /usr/local/share/jetson-utils/cmake/jetson-utilsConfig.cmake
-- Installing: /usr/local/share/jetson-utils/cmake/jetson-utilsConfig-noconfig.cmake
-- Installing: /usr/local/bin/camera-viewer
-- Set runtime path of "/usr/local/bin/camera-viewer" to ""
-- Installing: /usr/local/bin/gl-display-test
-- Set runtime path of "/usr/local/bin/gl-display-test" to ""
-- Installing: /usr/local/bin/camera-viewer.py
-- Installing: /usr/local/bin/cuda-from-numpy.py
-- Installing: /usr/local/bin/cuda-to-numpy.py
-- Installing: /usr/local/bin/gl-display-test.py
-- Installing: /usr/lib/python2.7/dist-packages/jetson_utils_python.so
-- Set runtime path of "/usr/lib/python2.7/dist-packages/jetson_utils_python.so" to ""
-- Installing: /usr/lib/python2.7/dist-packages/Jetson
-- Installing: /usr/lib/python2.7/dist-packages/Jetson/Utils
-- Installing: /usr/lib/python2.7/dist-packages/Jetson/Utils/__init__.py
-- Installing: /usr/lib/python2.7/dist-packages/Jetson/__init__.py
-- Installing: /usr/lib/python2.7/dist-packages/jetson
-- Installing: /usr/lib/python2.7/dist-packages/jetson/utils
-- Installing: /usr/lib/python2.7/dist-packages/jetson/utils/__init__.py
-- Installing: /usr/lib/python2.7/dist-packages/jetson/__init__.py
-- Installing: /usr/lib/python3.6/dist-packages/jetson_utils_python.so
-- Set runtime path of "/usr/lib/python3.6/dist-packages/jetson_utils_python.so" to ""
-- Installing: /usr/lib/python3.6/dist-packages/Jetson
-- Installing: /usr/lib/python3.6/dist-packages/Jetson/Utils
-- Installing: /usr/lib/python3.6/dist-packages/Jetson/Utils/__init__.py
-- Installing: /usr/lib/python3.6/dist-packages/Jetson/__init__.py
-- Installing: /usr/lib/python3.6/dist-packages/jetson
-- Installing: /usr/lib/python3.6/dist-packages/jetson/utils
-- Installing: /usr/lib/python3.6/dist-packages/jetson/utils/__init__.py
-- Installing: /usr/lib/python3.6/dist-packages/jetson/__init__.py
-- Installing: /usr/local/bin/detectnet-camera.py
-- Installing: /usr/local/bin/detectnet-console.py
-- Installing: /usr/local/bin/imagenet-camera.py
-- Installing: /usr/local/bin/imagenet-console.py
-- Installing: /usr/local/bin/my-detection.py
-- Installing: /usr/local/bin/my-recognition.py
-- Installing: /usr/local/bin/segnet-batch.py
-- Installing: /usr/local/bin/segnet-camera.py
-- Installing: /usr/local/bin/segnet-console.py
-- Installing: /usr/lib/python2.7/dist-packages/jetson_inference_python.so
-- Set runtime path of "/usr/lib/python2.7/dist-packages/jetson_inference_python.so" to ""
-- Up-to-date: /usr/lib/python2.7/dist-packages/Jetson
-- Installing: /usr/lib/python2.7/dist-packages/Jetson/__init__.py
-- Installing: /usr/lib/python2.7/dist-packages/Jetson/Inference
-- Installing: /usr/lib/python2.7/dist-packages/Jetson/Inference/__init__.py
-- Up-to-date: /usr/lib/python2.7/dist-packages/jetson
-- Installing: /usr/lib/python2.7/dist-packages/jetson/__init__.py
-- Installing: /usr/lib/python2.7/dist-packages/jetson/inference
-- Installing: /usr/lib/python2.7/dist-packages/jetson/inference/__init__.py
-- Installing: /usr/lib/python3.6/dist-packages/jetson_inference_python.so
-- Set runtime path of "/usr/lib/python3.6/dist-packages/jetson_inference_python.so" to ""
-- Up-to-date: /usr/lib/python3.6/dist-packages/Jetson
-- Installing: /usr/lib/python3.6/dist-packages/Jetson/__init__.py
-- Installing: /usr/lib/python3.6/dist-packages/Jetson/Inference
-- Installing: /usr/lib/python3.6/dist-packages/Jetson/Inference/__init__.py
-- Up-to-date: /usr/lib/python3.6/dist-packages/jetson
-- Installing: /usr/lib/python3.6/dist-packages/jetson/__init__.py
-- Installing: /usr/lib/python3.6/dist-packages/jetson/inference
-- Installing: /usr/lib/python3.6/dist-packages/jetson/inference/__init__.py

The project will be built to jetson-inference/build/aarch64, with the following directory structure:

|-build
  \aarch64
      \bin             where the sample binaries are built to
        \networks     where the network models are stored
        \images       where the test images are stored
      \include         where the headers reside
      \lib             where the libraries are build to

These also get installed under /usr/local/
The Python bindings for the jetson.inference and jetson.utils modules also get installed under /usr/lib/python*/dist-packages/.

view libjetson-utils and libjetson-inference in lib

$ tree build/aarch64/lib
.
├── libjetson-inference.so
├── libjetson-utils.so
└── python
    ├── 2.7
    │   ├── jetson_inference_python.so
    │   └── jetson_utils_python.so
    └── 3.6
        ├── jetson_inference_python.so
        └── jetson_utils_python.so

3 directories, 6 files

libjetson-inference.so

$ ldd libjetson-inference.so 

    linux-vdso.so.1 (0x0000007fa6e8c000)
    libpthread.so.0 => /lib/aarch64-linux-gnu/libpthread.so.0 (0x0000007fa6cf3000)
    libdl.so.2 => /lib/aarch64-linux-gnu/libdl.so.2 (0x0000007fa6cde000)
    librt.so.1 => /lib/aarch64-linux-gnu/librt.so.1 (0x0000007fa6cc7000)
    libjetson-utils.so => /home/nano/git/jetson-inference/build/aarch64/lib/libjetson-utils.so (0x0000007fa6b6f000)
    libnvinfer.so.5 => /usr/lib/aarch64-linux-gnu/libnvinfer.so.5 (0x0000007f9dc23000)
    libnvinfer_plugin.so.5 => /usr/lib/aarch64-linux-gnu/libnvinfer_plugin.so.5 (0x0000007f9d94d000)
    libnvparsers.so.5 => /usr/lib/aarch64-linux-gnu/libnvparsers.so.5 (0x0000007f9d60e000)
    libnvonnxparser.so.0 => /usr/lib/aarch64-linux-gnu/libnvonnxparser.so.0 (0x0000007f9d1ea000)
    libopencv_calib3d.so.3.3 => /usr/lib/libopencv_calib3d.so.3.3 (0x0000007f9d0be000)
    libopencv_core.so.3.3 => /usr/lib/libopencv_core.so.3.3 (0x0000007f9cde9000)
    libstdc++.so.6 => /usr/lib/aarch64-linux-gnu/libstdc++.so.6 (0x0000007f9cc56000)
    libm.so.6 => /lib/aarch64-linux-gnu/libm.so.6 (0x0000007f9cb9c000)
    libgcc_s.so.1 => /lib/aarch64-linux-gnu/libgcc_s.so.1 (0x0000007f9cb78000)
    libc.so.6 => /lib/aarch64-linux-gnu/libc.so.6 (0x0000007f9ca1f000)
    /lib/ld-linux-aarch64.so.1 (0x0000007fa6e61000)
    libGL.so.1 => /usr/lib/aarch64-linux-gnu/libGL.so.1 (0x0000007f9c920000)
    libGLEW.so.2.0 => /usr/lib/aarch64-linux-gnu/libGLEW.so.2.0 (0x0000007f9c874000)
    libgstreamer-1.0.so.0 => /usr/lib/aarch64-linux-gnu/libgstreamer-1.0.so.0 (0x0000007f9c744000)
    libgstapp-1.0.so.0 => /usr/lib/aarch64-linux-gnu/libgstapp-1.0.so.0 (0x0000007f9c726000)
    libcudnn.so.7 => /usr/lib/aarch64-linux-gnu/libcudnn.so.7 (0x0000007f858c0000)
    libcublas.so.10.0 => /usr/local/cuda-10.0/targets/aarch64-linux/lib/libcublas.so.10.0 (0x0000007f7ff59000)
    libcudart.so.10.0 => /usr/local/cuda-10.0/targets/aarch64-linux/lib/libcudart.so.10.0 (0x0000007f7fee8000)
    libopencv_flann.so.3.3 => /usr/lib/libopencv_flann.so.3.3 (0x0000007f7fe85000)
    libopencv_imgproc.so.3.3 => /usr/lib/libopencv_imgproc.so.3.3 (0x0000007f7f6b8000)

imageNet demo

C++

$ cd etson-inference/build/aarch64/bin
$ sudo ./imagenet-console --network=resnet-18 images/orange_0.jpg output_0.jpg

output

imageNet -- loading classification network model from:
         -- prototxt     networks/ResNet-18/deploy.prototxt
         -- model        networks/ResNet-18/ResNet-18.caffemodel
         -- class_labels networks/ilsvrc12_synset_words.txt
         -- input_blob   'data'
         -- output_blob  'prob'
         -- batch_size   1

[TRT]   TensorRT version 5.1.6
[TRT]   loading NVIDIA plugins...
[TRT]   Plugin Creator registration succeeded - GridAnchor_TRT
[TRT]   Plugin Creator registration succeeded - NMS_TRT
[TRT]   Plugin Creator registration succeeded - Reorg_TRT
[TRT]   Plugin Creator registration succeeded - Region_TRT
[TRT]   Plugin Creator registration succeeded - Clip_TRT
[TRT]   Plugin Creator registration succeeded - LReLU_TRT
[TRT]   Plugin Creator registration succeeded - PriorBox_TRT
[TRT]   Plugin Creator registration succeeded - Normalize_TRT
[TRT]   Plugin Creator registration succeeded - RPROI_TRT
[TRT]   Plugin Creator registration succeeded - BatchedNMS_TRT
[TRT]   completed loading NVIDIA plugins.
[TRT]   detected model format - caffe  (extension '.caffemodel')
[TRT]   desired precision specified for GPU: FASTEST
[TRT]   requested fasted precision for device GPU without providing valid calibrator, disabling INT8
[TRT]   native precisions detected for GPU:  FP32, FP16
[TRT]   selecting fastest native precision for GPU:  FP16
[TRT]   attempting to open engine cache file networks/ResNet-18/ResNet-18.caffemodel.1.1.GPU.FP16.engine
[TRT]   cache file not found, profiling network model on device GPU
[TRT]   device GPU, loading networks/ResNet-18/deploy.prototxt networks/ResNet-18/ResNet-18.caffemodel
[TRT]   retrieved Output tensor "prob":  1000x1x1
[TRT]   retrieved Input tensor "data":  3x224x224
[TRT]   device GPU, configuring CUDA engine
[TRT]   device GPU, building FP16:  ON
[TRT]   device GPU, building INT8:  OFF
[TRT]   device GPU, building CUDA engine (this may take a few minutes the first time a network is loaded)
[TRT]   device GPU, completed building CUDA engine
[TRT]   network profiling complete, writing engine cache to networks/ResNet-18/ResNet-18.caffemodel.1.1.GPU.FP16.engine
[TRT]   device GPU, completed writing engine cache to networks/ResNet-18/ResNet-18.caffemodel.1.1.GPU.FP16.engine
[TRT]   device GPU, networks/ResNet-18/ResNet-18.caffemodel loaded
[TRT]   device GPU, CUDA engine context initialized with 2 bindings
[TRT]   binding -- index   0
               -- name    'data'
               -- type    FP32
               -- in/out  INPUT
               -- # dims  3
               -- dim #0  3 (CHANNEL)
               -- dim #1  224 (SPATIAL)
               -- dim #2  224 (SPATIAL)
[TRT]   binding -- index   1
               -- name    'prob'
               -- type    FP32
               -- in/out  OUTPUT
               -- # dims  3
               -- dim #0  1000 (CHANNEL)
               -- dim #1  1 (SPATIAL)
               -- dim #2  1 (SPATIAL)
[TRT]   binding to input 0 data  binding index:  0
[TRT]   binding to input 0 data  dims (b=1 c=3 h=224 w=224) size=602112
[TRT]   binding to output 0 prob  binding index:  1
[TRT]   binding to output 0 prob  dims (b=1 c=1000 h=1 w=1) size=4000
device GPU, networks/ResNet-18/ResNet-18.caffemodel initialized.
[TRT]   networks/ResNet-18/ResNet-18.caffemodel loaded
imageNet -- loaded 1000 class info entries
networks/ResNet-18/ResNet-18.caffemodel initialized.
[image] loaded 'images/orange_0.jpg'  (1920 x 1920, 3 channels)
class 0950 - 0.996028  (orange)
imagenet-console:  'images/orange_0.jpg' -> 99.60276% class #950 (orange)

[TRT]   ------------------------------------------------
[TRT]   Timing Report networks/ResNet-18/ResNet-18.caffemodel
[TRT]   ------------------------------------------------
[TRT]   Pre-Process   CPU   0.10824ms  CUDA   0.34156ms
[TRT]   Network       CPU  12.91854ms  CUDA  12.47026ms
[TRT]   Post-Process  CPU   0.80311ms  CUDA   0.82672ms
[TRT]   Total         CPU  13.82989ms  CUDA  13.63854ms
[TRT]   ------------------------------------------------

[TRT]   note -- when processing a single image, run 'sudo jetson_clocks' before
                to disable DVFS for more accurate profiling/timing measurements

imagenet-console:  attempting to save output image to 'output_0.jpg'
imagenet-console:  completed saving 'output_0.jpg'
imagenet-console:  shutting down...
imagenet-console:  shutdown complete

Python

$ cd etson-inference/build/aarch64/bin
$ sudo ./imagenet-console.py --network=resnet-18 images/orange_0.jpg output_0.jpg

output

jetson.inference.__init__.py
jetson.inference -- initializing Python 2.7 bindings...
jetson.inference -- registering module types...
jetson.inference -- done registering module types
jetson.inference -- done Python 2.7 binding initialization
jetson.utils.__init__.py
jetson.utils -- initializing Python 2.7 bindings...
jetson.utils -- registering module functions...
jetson.utils -- done registering module functions
jetson.utils -- registering module types...
jetson.utils -- done registering module types
jetson.utils -- done Python 2.7 binding initialization
[image] loaded 'images/orange_0.jpg'  (1920 x 1920, 3 channels)
jetson.inference -- PyTensorNet_New()
jetson.inference -- PyImageNet_Init()
jetson.inference -- imageNet loading network using argv command line params
jetson.inference -- imageNet.__init__() argv[0] = './imagenet-console.py'
jetson.inference -- imageNet.__init__() argv[1] = '--network=resnet-18'
jetson.inference -- imageNet.__init__() argv[2] = 'images/orange_0.jpg'
jetson.inference -- imageNet.__init__() argv[3] = 'output_0.jpg'

imageNet -- loading classification network model from:
         -- prototxt     networks/ResNet-18/deploy.prototxt
         -- model        networks/ResNet-18/ResNet-18.caffemodel
         -- class_labels networks/ilsvrc12_synset_words.txt
         -- input_blob   'data'
         -- output_blob  'prob'
         -- batch_size   1

[TRT]   TensorRT version 5.1.6
[TRT]   loading NVIDIA plugins...
[TRT]   Plugin Creator registration succeeded - GridAnchor_TRT
[TRT]   Plugin Creator registration succeeded - NMS_TRT
[TRT]   Plugin Creator registration succeeded - Reorg_TRT
[TRT]   Plugin Creator registration succeeded - Region_TRT
[TRT]   Plugin Creator registration succeeded - Clip_TRT
[TRT]   Plugin Creator registration succeeded - LReLU_TRT
[TRT]   Plugin Creator registration succeeded - PriorBox_TRT
[TRT]   Plugin Creator registration succeeded - Normalize_TRT
[TRT]   Plugin Creator registration succeeded - RPROI_TRT
[TRT]   Plugin Creator registration succeeded - BatchedNMS_TRT
[TRT]   completed loading NVIDIA plugins.
[TRT]   detected model format - caffe  (extension '.caffemodel')
[TRT]   desired precision specified for GPU: FASTEST
[TRT]   requested fasted precision for device GPU without providing valid calibrator, disabling INT8
[TRT]   native precisions detected for GPU:  FP32, FP16
[TRT]   selecting fastest native precision for GPU:  FP16
[TRT]   attempting to open engine cache file networks/ResNet-18/ResNet-18.caffemodel.1.1.GPU.FP16.engine
[TRT]   loading network profile from engine cache... networks/ResNet-18/ResNet-18.caffemodel.1.1.GPU.FP16.engine
[TRT]   device GPU, networks/ResNet-18/ResNet-18.caffemodel loaded
[TRT]   device GPU, CUDA engine context initialized with 2 bindings
[TRT]   binding -- index   0
               -- name    'data'
               -- type    FP32
               -- in/out  INPUT
               -- # dims  3
               -- dim #0  3 (CHANNEL)
               -- dim #1  224 (SPATIAL)
               -- dim #2  224 (SPATIAL)
[TRT]   binding -- index   1
               -- name    'prob'
               -- type    FP32
               -- in/out  OUTPUT
               -- # dims  3
               -- dim #0  1000 (CHANNEL)
               -- dim #1  1 (SPATIAL)
               -- dim #2  1 (SPATIAL)
[TRT]   binding to input 0 data  binding index:  0
[TRT]   binding to input 0 data  dims (b=1 c=3 h=224 w=224) size=602112
[TRT]   binding to output 0 prob  binding index:  1
[TRT]   binding to output 0 prob  dims (b=1 c=1000 h=1 w=1) size=4000
device GPU, networks/ResNet-18/ResNet-18.caffemodel initialized.
[TRT]   networks/ResNet-18/ResNet-18.caffemodel loaded
imageNet -- loaded 1000 class info entries
networks/ResNet-18/ResNet-18.caffemodel initialized.
class 0950 - 0.996028  (orange)
image is recognized as 'orange' (class #950) with 99.602759% confidence


[TRT]   ------------------------------------------------
[TRT]   Timing Report networks/ResNet-18/ResNet-18.caffemodel
[TRT]   ------------------------------------------------
[TRT]   Pre-Process   CPU   0.06884ms  CUDA   0.32849ms
[TRT]   Network       CPU  11.44888ms  CUDA  11.01536ms
[TRT]   Post-Process  CPU   0.20783ms  CUDA   0.20708ms
[TRT]   Total         CPU  11.72555ms  CUDA  11.55094ms
[TRT]   ------------------------------------------------

[TRT]   note -- when processing a single image, run 'sudo jetson_clocks' before
                to disable DVFS for more accurate profiling/timing measurements

jetson.utils -- PyFont_New()
jetson.utils -- PyFont_Init()
jetson.utils -- PyFont_Dealloc()
jetson.utils -- freeing CUDA mapped memory
PyTensorNet_Dealloc()

my-recognition

# build
$ cd jetson-inference/examples/my-recognition
$ mkdir build && cd build && cmake-gui ..

my-recognition cmake-gui

# compile 
$ make 

Scanning dependencies of target my-recognition
[ 50%] Building CXX object CMakeFiles/my-recognition.dir/my-recognition.cpp.o
[100%] Linking CXX executable my-recognition
[100%] Built target my-recognition

# view libraries
$ ldd my-recognition 
    linux-vdso.so.1 (0x0000007fb5546000)
    libjetson-inference.so => /usr/local/lib/libjetson-inference.so (0x0000007fb53ea000)
    libjetson-utils.so => /usr/local/lib/libjetson-utils.so (0x0000007fb5292000)
    libstdc++.so.6 => /usr/lib/aarch64-linux-gnu/libstdc++.so.6 (0x0000007fb50db000)
    libc.so.6 => /lib/aarch64-linux-gnu/libc.so.6 (0x0000007fb4f82000)
    /lib/ld-linux-aarch64.so.1 (0x0000007fb551b000)
    libpthread.so.0 => /lib/aarch64-linux-gnu/libpthread.so.0 (0x0000007fb4f56000)
    libdl.so.2 => /lib/aarch64-linux-gnu/libdl.so.2 (0x0000007fb4f41000)
    librt.so.1 => /lib/aarch64-linux-gnu/librt.so.1 (0x0000007fb4f2a000)
    libnvinfer.so.5 => /usr/lib/aarch64-linux-gnu/libnvinfer.so.5 (0x0000007fabfde000)
    libnvinfer_plugin.so.5 => /usr/lib/aarch64-linux-gnu/libnvinfer_plugin.so.5 (0x0000007fabd08000)
    libnvparsers.so.5 => /usr/lib/aarch64-linux-gnu/libnvparsers.so.5 (0x0000007fab9c9000)
    libnvonnxparser.so.0 => /usr/lib/aarch64-linux-gnu/libnvonnxparser.so.0 (0x0000007fab5a5000)
    libopencv_calib3d.so.3.3 => /usr/lib/libopencv_calib3d.so.3.3 (0x0000007fab479000)
    libopencv_core.so.3.3 => /usr/lib/libopencv_core.so.3.3 (0x0000007fab1a4000)

run and get result

$ ./build/my-recognition polar_bear.jpg

class 0296 - 0.997434  (ice bear, polar bear, Ursus Maritimus, Thalarctos maritimus)

imagenet-camera

imagenet-camera usage

$ ./imagenet-camera --help
usage: imagenet-camera [-h] [--network NETWORK] [--camera CAMERA]
                       [--width WIDTH] [--height HEIGHT]

Classify a live camera stream using an image recognition DNN.

optional arguments:
  --help            show this help message and exit
  --network NETWORK pre-trained model to load (see below for options)
  --camera CAMERA   index of the MIPI CSI camera to use (e.g. CSI camera 0),
                    or for VL42 cameras, the /dev/video device to use.
                    by default, MIPI CSI camera 0 will be used.
  --width WIDTH     desired width of camera stream (default is 1280 pixels)
  --height HEIGHT   desired height of camera stream (default is 720 pixels)

imageNet arguments: 
  --network NETWORK    pre-trained model to load, one of the following:
                           * alexnet
                           * googlenet (default)
                           * googlenet-12
                           * resnet-18
                           * resnet-50
                           * resnet-101
                           * resnet-152
                           * vgg-16
                           * vgg-19
                           * inception-v4
  --model MODEL        path to custom model to load (caffemodel, uff, or onnx)
  --prototxt PROTOTXT  path to custom prototxt to load (for .caffemodel only)
  --labels LABELS      path to text file containing the labels for each class
  --input_blob INPUT   name of the input layer (default is 'data')
  --output_blob OUTPUT name of the output layer (default is 'prob')
  --batch_size BATCH   maximum batch size (default is 1)
  --profile            enable layer profiling in TensorRT

camera type

  • MIPI CSI cameras are used by specifying the sensor index (0 or 1, ect.)
  • V4L2 USB cameras are used by specifying their /dev/video node (/dev/video0, /dev/video1, ect.)

    The default is to use MIPI CSI sensor 0 (--camera=0)

Query the available formats with the following commands:

$ sudo apt-get install -y v4l-utils
$ v4l2-ctl --list-formats-ext

ioctl: VIDIOC_ENUM_FMT
    Index       : 0
    Type        : Video Capture
    Pixel Format: 'MJPG' (compressed)
    Name        : Motion-JPEG
        Size: Discrete 1920x1080
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 160x120
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 176x144
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 320x240
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 352x288
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 640x360
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 640x480
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 1280x720
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 1280x1024
            Interval: Discrete 0.033s (30.000 fps)

    Index       : 1
    Type        : Video Capture
    Pixel Format: 'YUYV'
    Name        : YUYV 4:2:2
        Size: Discrete 1920x1080
            Interval: Discrete 0.200s (5.000 fps)
        Size: Discrete 160x120
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 176x144
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 320x240
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 352x288
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 640x360
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 640x480
            Interval: Discrete 0.033s (30.000 fps)
        Size: Discrete 1280x720
            Interval: Discrete 0.100s (10.000 fps)
        Size: Discrete 1280x1024
            Interval: Discrete 0.200s (5.000 fps)

run demo

$ ./imagenet-camera --network=resnet-18  --camera=0 --width=640 --height=480

nvpmodel

# /etc/nvpmodel.conf

sudo nvpmodel -q –-verbose # 查看当前的模式
sudo nvpmodel -p –-verbose # 打印支持的所有模式及其配置
sudo nvpmodel -m 0 # 启动最高性能,此时所有CPU均已启动,但对应的主频还不是最高的
sudo ~/jetson_clocks.sh # 开启最大频率

Reference

History

  • 2019/12/09: created.

Author: kezunlin
Reprint policy: All articles in this blog are used except for special statements CC BY 4.0 reprint polocy. If reproduced, please indicate source kezunlin !
评论
  TOC