Skip to content

Error_orcuring_when_install_jetson_nano.md

Latest commit

 

History

History
executable file
·
987 lines (709 loc) · 30.3 KB

Error_orcuring_when_install_jetson_nano.md

File metadata and controls

executable file
·
987 lines (709 loc) · 30.3 KB

[NOTE] Jetson Download Center

Reference:

https://developer.nvidia.com/embedded/downloads

[ERROR] N: Unable to locate package python-pip

Reason:

have to enable universe category which contains python-pip package.

Solution:

01. sudo apt-get update
02. sudo apt-get upgrade
03. sudo apt install nano -y 
04. sudo apt-get install htop -y
05. sudo apt-get install tmux -y
05. sudo apt install python3-pip -y
06. sudo -H pip3 install jetson-stats
07. sudo -H pip3 install --upgrade setuptools
08. sudo -H pip3 install cython
09. sudo -H pip3 install -U setuptools
10. sudo nano /etc/apt/sources.list
11. add universe at the end of file

# See http://help.ubuntu.com/community/UpgradeNotes for how to upgrade to
# newer versions of the distribution.
deb http://ports.ubuntu.com/ubuntu-ports/ bionic main restricted universe
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic main restricted

## Major bug fix updates produced after the final release of the
## distribution.
deb http://ports.ubuntu.com/ubuntu-ports/ bionic-updates main restricted universe
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic-updates main restricted

## N.B. software from this repository is ENTIRELY UNSUPPORTED by the Ubuntu
## team. Also, please note that software in universe WILL NOT receive any
## review or updates from the Ubuntu security team.
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic universe
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic-updates universe

## N.B. software from this repository is ENTIRELY UNSUPPORTED by the Ubuntu
## team, and may not be under a free licence. Please satisfy yourself as to
## your rights to use the software. Also, please note that software in
## multiverse WILL NOT receive any review or updates from the Ubuntu
## security team.
deb http://ports.ubuntu.com/ubuntu-ports/ bionic multiverse
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic multiverse
deb http://ports.ubuntu.com/ubuntu-ports/ bionic-updates multiverse
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic-updates multiverse

## N.B. software from this repository may not have been tested as
## extensively as that contained in the main release, although it includes
## newer versions of some applications which may provide useful features.
## Also, please note that software in backports WILL NOT receive any review
## or updates from the Ubuntu security team.
deb http://ports.ubuntu.com/ubuntu-ports/ bionic-backports main restricted multiverse universe
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic-backports main restricted universe multiverse

## Uncomment the following two lines to add software from Canonical's
## 'partner' repository.##
## This software is not part of Ubuntu, but is offered by Canonical and the
## respective vendors as a service to Ubuntu users.
deb http://archive.canonical.com/ubuntu bionic partner
# deb-src http://archive.canonical.com/ubuntu bionic partner

deb http://ports.ubuntu.com/ubuntu-ports/ bionic-security main restricted universe
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic-security main restricted
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic-security universe
deb http://ports.ubuntu.com/ubuntu-ports/ bionic-security multiverse
# deb-src http://ports.ubuntu.com/ubuntu-ports/ bionic-security multiverse

[ERROR] The repository does not have a Release file

Run sudo apt update and get error:

E: The repository 'http://ppa.launchpad.net/armagetronad-dev/ppa/ubuntu bionic Release' does not have a Release file. 
N: Updating from such a repository can't be done securely, and is therefore disabled by default. 
N: See apt-secure(8) manpage for repository creation and user configuration details.

Solution:

sudo apt-add-repository -r ppa:armagetronad-dev/ppa
sudo apt update -q

[ERROR] E: Unable to locate package nano

Solution:

snap install nano

[TUITORIAL] Check available camera

ls -ltrh /dev/video*

[TUITORIAL] Jetson Nano CSI-Camera

git clone https://github.com/JetsonHacksNano/CSI-Camera.git
git checkout v2.0

[ERROR] W: GPG error

W: GPG error: http://archive.debian.org/debian jessie-backports InRelease: The following signatures couldn't be verified because the public key is not avail able: NO_PUBKEY 8B48AD6246925553 NO_PUBKEY 7638D0442B90D010

Reason:

apt is telling us that the following keys are missing: 8B48AD6246925553 and 7638D0442B90D010

Solution:

sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 8B48AD6246925553
sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 7638D0442B90D010

[ERROR] E: Release file for .../InRelease is expired

E: Release file for http://archive.debian.org/debian/dists/jessie-backports/InRelease is expired (invalid since 357d 12h 49min 26s). Updates for this repository will not be applied. Solution:

add "-o Acquire::Check-Valid-Until=false" into the end of command.
eg: sudo apt update -o Acquire::Check-Valid-Until=false

Setup VNC server on the Jetson developer kit

REference

  • Enable the VNC server to start each time you log in
pushd /usr/lib/systemd/user/graphical-session.target.wants
sudo ln -s ../vino-server.service ./.
  • Configure the VNC server
gsettings set org.gnome.Vino prompt-enabled false
gsettings set org.gnome.Vino require-encryption false
  • Set a password to access the VNC server
# Replace thepassword with your desired password
gsettings set org.gnome.Vino authentication-methods "['vnc']"
gsettings set org.gnome.Vino vnc-password $(echo -n 'thepassword'|base64)
  • Edit file config
sudo nano /usr/share/glib-2.0/schemas/org.gnome.Vino.gschema.xml
    <key name='enabled' type='b'>
      <summary>Enable remote access to the desktop</summary>
      <description>
        If true, allows remote access to the desktop via the RFB
        protocol. Users on remote machines may then connect to the
        desktop using a VNC viewer.
      </description>
      <default>false</default>
    </key>
  • recompile
sudo glib-compile-schemas /usr/share/glib-2.0/schemas
  • Create pseudo screen
sudo nano /etc/X11/xorg.conf
Section "Screen"
    Identifier "Default Screen"
    Monitor "Configured Monitor"
    Device "Tegra0"
    DefaultDepth 24
    SubSection "Display"
        Depth 24
        Virtual 1920 1080
    EndSubSection
EndSection
  • Reboot the system so that the settings take effect
sudo reboot

/etc/sudoers file corrupted and I can't run 'pkexec visudo' over SSH

  • Open two ssh sessions to the target server.

  • In the first session, get the PID of bash by running: echo $$

  • In the second session, start the authentication agent with: pkttyagent --process (pid from step 2)

  • Back in the first session, run: pkexec visudo

  • In the second session, you will get the password prompt. visudo will start in the first session.

[TUITORIAL] Uninstall OpenCV Completely

Reference

https://installlion.com/ubuntu/xenial/universe/p/python-opencv/uninstall/index.html

Uninstall just python-opencv

sudo apt-get remove python-opencv

Uninstall python-opencv and its dependencies

sudo apt-get remove --auto-remove python-opencv

Purging your config/data too

sudo apt-get purge --auto-remove python-opencv

[TUITORIAL] Install OpenCV 3.4.6

[ ] Reference:

https://github.com/jkjung-avt/jetson_nano.git

Step 0: Add Swap Memory

sudo fallocate -l 4G /var/swapfile
sudo chmod 600 /var/swapfile
sudo mkswap /var/swapfile
sudo swapon /var/swapfile
sudo bash -c 'echo "/var/swapfile swap swap defaults 0 0" >> /etc/fstab'

Step 1: Install OpenCV dependencies on Ubuntu 18.04

sudo apt-get update
sudo apt-get upgrade -y
sudo apt-get install build-essential -y
sudo apt-get install cmake -y
sudo apt-get install unzip -y
sudo apt-get install pkg-config -y
sudo apt-get install libjpeg-dev -y
sudo apt-get install libpng-dev -y
sudo apt-get install libtiff-dev -y
sudo apt-get install libjasper-dev -y
sudo apt-get install libavcodec-dev -y
sudo apt-get install libavformat-dev -y
sudo apt-get install libswscale-dev -y
sudo apt-get install libv4l-dev -y
sudo apt-get install libxvidcore-dev -y
sudo apt-get install libx264-dev -y
sudo apt-get install libgtk-3-dev -y
sudo apt-get install libatlas-base-dev -y
sudo apt-get install gfortran -y
sudo apt-get install python3.6-dev -y

Step 2: Download the official OpenCV source:

cd ~
wget -O opencv.zip https://github.com/opencv/opencv/archive/3.4.6.zip
wget -O opencv_contrib.zip https://github.com/opencv/opencv_contrib/archive/3.4.6.zip
unzip opencv.zip
unzip opencv_contrib.zip
mv opencv-3.4.6 opencv
mv opencv_contrib-3.4.6 opencv_contrib
pip3 install numpy
cd ~/opencv
mkdir build
cd build
cmake \
    -D CMAKE_BUILD_TYPE=RELEASE \
    -D CMAKE_INSTALL_PREFIX=/usr/local \
    -D WITH_CUDA=ON \
    -D INSTALL_PYTHON_EXAMPLES=ON \
    -D OPENCV_EXTRA_MODULES_PATH=~/opencv_contrib/modules \
    -D OPENCV_ENABLE_NONFREE=ON \
    -D BUILD_EXAMPLES=ON \
    -D WITH_GSTREAMER=ON ..

sudo make -j3
# sudo make -j$(nproc)

sudo make install
sudo ldconfig
pkg-config --modversion opencv

Step 3: Find & Set “opencv.pc” file path:

ls /usr/local/lib/pkgconfig/
sudo cp /usr/local/lib/pkgconfig/opencv.pc  /usr/lib/x86_64-linux-gnu/pkgconfig/opencv.pc
pkg-config --modversion opencv

Step 4: Compile code:

code

.
├── build
├── CMakeLists.txt
├── main.cpp
└── makefile

Makefile

GCC ?= g++

CCFLAGS := -std=c++11

LDFLAGS += $(shell pkg-config --libs opencv)
INCLUDES += $(shell pkg-config --cflags opencv)

SRCS=main.cpp
OBJS=main.o
RM=rm -f
EXEC=app

# Target rules
all: build

build: ${EXEC}

%.o: %.cpp %.h
	$(GCC) $(CCFLAGS) $(INCLUDES) -o $@ -c $<

${EXEC}: ${OBJS}
	$(GCC) $(CCFLAGS) -o $@ $^ $(LDFLAGS)

clean:
	${RM} ${EXEC} *.o 

dist-clean:
	$(RM) $(EXEC)

CMakeLists.txt

cmake_minimum_required(VERSION 2.8)
project( HNIW )

find_package( OpenCV REQUIRED )
include_directories( ${OpenCV_INCLUDE_DIRS} )

add_executable( app main.cpp )
target_link_libraries( app ${OpenCV_LIBS} )

[TUITORIAL] Install OpenCV 4

Add Swap Memory

sudo fallocate -l 4G /var/swapfile
sudo chmod 600 /var/swapfile
sudo mkswap /var/swapfile
sudo swapon /var/swapfile
sudo bash -c 'echo "/var/swapfile swap swap defaults 0 0" >> /etc/fstab'

Updating the packages

link to Long_senpai

Download OpenCV & Contribs Modules

pushd ~/
git clone https://github.com/opencv/opencv.git
git clone https://github.com/opencv/opencv_contrib.git

Create directory

mkdir build
cd build/

Build Opencv using Cmake

cmake -D CMAKE_BUILD_TYPE=RELEASE \
    -D CMAKE_C_COMPILER=/usr/bin/gcc \
    -D CMAKE_INSTALL_PREFIX=/usr/local \
    -D INSTALL_C_EXAMPLES=ON \
    -D WITH_TBB=ON \
    -D WITH_CUDA=ON \
    -D WITH_CUDNN=ON \
    -D OPENCV_DNN_CUDA=ON \
    -D BUILD_opencv_cudacodec=ON \
    -D ENABLE_FAST_MATH=1 \
    -D CUDA_FAST_MATH=1 \
    -D WITH_CUBLAS=1 \
    -D WITH_V4L=ON \
    -D WITH_QT=OFF \
    -D WITH_OPENGL=ON \
    -D WITH_GSTREAMER=ON \
    -D OPENCV_GENERATE_PKGCONFIG=ON \
    -D OPENCV_PC_FILE_NAME=opencv.pc \
    -D OPENCV_ENABLE_NONFREE=ON \
    -D OPENCV_EXTRA_MODULES_PATH=~/opencv_contrib/modules \
    -D BUILD_SHARED_LIBS=ON \
    -D WITH_FFMPEG=ON \
    -D WITH_OPENCL=ON \
    -D BUILD_EXAMPLES=ON \
    -D INSTALL_PYTHON_EXAMPLES=ON \
    -D PYTHON_EXECUTABLE=/usr/bin/python3 \
    -D OPENCV_PYTHON3_INSTALL_PATH=/usr/local/lib/python3.8/dist-packages/ \
    -D BUILD_TESTS=OFF \
    -D BUILD_PERF_TESTS=OFF ..

Compile the OpenCV with Contribs Modules

make -j$(nproc)
sudo make install

Find & Set “opencv.pc” file path

Compile code

Install PyCuda

  • By pip
python -m pip install pycuda
  • Build
arch=$(uname -m)
folder=${HOME}/src
mkdir -p $folder
sudo apt-get install -y build-essential python3-dev
sudo apt-get install -y libboost-python-dev libboost-thread-dev
boost_pylib=$(basename /usr/lib/${arch}-linux-gnu/libboost_python*-py3?.so)
boost_pylibname=${boost_pylib%.so}
boost_pyname=${boost_pylibname/lib/}
pushd $folder
wget https://files.pythonhosted.org/packages/5a/56/4682a5118a234d15aa1c8768a528aac4858c7b04d2674e18d586d3dfda04/pycuda-2021.1.tar.gz
CPU_CORES=$(nproc)
tar xzvf pycuda-2021.1.tar.gz
cd pycuda-2021.1
cat ./setup.py | sed 's/"nvcc"/"\/usr\/local\/cuda\/bin\/nvcc"/g' > m_setup.py
mv m_setup.py setup.py
sudo $(which python) ./configure.py --python-exe=$(which python) --cuda-root=/usr/local/cuda --cudadrv-lib-dir=/usr/lib/${arch}-linux-gnu --boost-inc-dir=/usr/include --boost-lib-dir=/usr/lib/${arch}-linux-gnu --boost-python-libname=${boost_pyname} --boost-thread-libname=boost_thread
make -j$CPU_CORES
sudo $(which python) setup.py build
sudo $(which python) setup.py install
popd
sudo $(which python) -c "import pycuda; print('pycuda version:', pycuda.VERSION)"

[TUITORIAL] Install Tensorflow

[ ]Reference:

https://forums.developer.nvidia.com/t/official-tensorflow-for-jetson-nano/71770

Python 3.6 + JetPack 4.5

$ sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev liblapack-dev libblas-dev gfortran
$ sudo apt-get install python3-pip
$ sudo pip3 install -U pip testresources setuptools==49.6.0
$ sudo pip3 install -U numpy==1.16.1 future==0.18.2 mock==3.0.5 h5py==2.10.0 keras_preprocessing==1.1.1 keras_applications==1.0.8 gast==0.2.2 futures protobuf pybind11

# TF-2.x
$ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v45 tensorflow

# TF-1.15
$ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v45 'tensorflow<2'

Python 3.6 + JetPack 4.4

$ sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev liblapack-dev libblas-dev gfortran
$ sudo apt-get install python3-pip
$ sudo pip3 install -U pip
$ sudo pip3 install -U pip testresources setuptools numpy==1.16.1 future==0.17.1 mock==3.0.5 h5py==2.9.0 keras_preprocessing==1.0.5 keras_applications==1.0.8 gast==0.2.2 futures protobuf pybind11

# TF-2.x
$ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v44 tensorflow==2.3.1+nv20.12

# TF-1.15
$ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v44 'tensorflow<2'

Python 3.6 + JetPack 4.3

$ sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev
$ sudo apt-get install python3-pip
$ sudo pip3 install -U pip
$ sudo pip3 install -U numpy grpcio absl-py py-cpuinfo psutil portpicker six mock requests gast h5py astor termcolor protobuf keras-applications keras-preprocessing wrapt google-pasta

# TF-2.x
$ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v43 tensorflow==2.1.0+nv20.3

# TF-1.15
$ sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v43 tensorflow==1.15.2+nv20.3

Note

If there is some error when import tensorflow after installation, run export OPENBLAS_CORETYPE=ARMV8 before run python app.

[TUITORIAL] Pytorch v1.4.0 for Jetson Nano

Pytorch v1.4.0 for Jetson Nano

[ ] Reference:

https://forums.developer.nvidia.com/t/pytorch-for-jetson-nano-version-1-4-0-now-available/72048

1

wget https://nvidia.box.com/shared/static/ncgzus5o23uck9i5oth2n8n06k340l6k.whl -O torch-1.4.0-cp36-cp36m-linux_aarch64.whl

2

sudo apt-get install python3-pip libopenblas-base

3

pip3 install Cython

4

pip3 install numpy torch-1.4.0-cp36-cp36m-linux_aarch64.whl

Pytorch v1.4.0 for Jetson Nano

[ ] Reference:

https://medium.com/hackers-terminal/installing-pytorch-torchvision-on-nvidias-jetson-tx2-81591d03ce32

1

sudo apt-get install libjpeg-dev zlib1g-dev

2 git clone -b v0.5.0 https://github.com/pytorch/vision torchvision

3

cd torchvision

4

sudo python3 setup.py install

Version of torchvision to download depending on the version of PyTorch

Pytorch torchvision
v1.0 v0.2.2
v1.1 v0.3.0
v1.2 v0.4.0
v1.3 v0.4.2
v1.4 v0.5.0

[TUITORIAL] Generating TensorRT Engines from TensorFlow

[ ] Reference: https://docs.nvidia.com/deeplearning/sdk/tensorrt-archived/tensorrt_401/tensorrt-api/python_api/index.html

1. Check TensorRT version (the current TensorRT version is 6.0.1) dpkg -l | grep nvinfer

https://blog.csdn.net/ctwy291314/article/details/103507846

2 Check CUDA version

cat /usr/local/cuda/version.txt

3 Check CUDNN version

cat /usr/include/aarch64-linux-gnu/cudnn_v7.h | grep CUDNN_MAJOR -A 2

4 Installing the Python APIs:

pip3 install tensorrt-*_16.04-cp35-cp35m-linux_x86_64.whl

[TUITORIAL] Install Jupiter Lab

sudo apt install python3-pip
sudo pip3 install wheel
sudo pip3 install jupyter jupyterlab
sudo jupyter labextension install @jupyter-widgets/jupyterlab-manager
jupyter lab --generate-config

jupyter notebook remote

jupyter notebook --no-browser --port=8889
ssh -N -f -L 8888:8889 jetson@192.168.137.189
http://localhost:8888/?token=a9dca23be19c7a142c5d593d9c717c81354fd1062ab90fad

Add Conda environment to jupyter notebook

python -m ipykernel install --user --name M --display-name "MINH"

[TUITORIAL] Setup Anaconda

[ ] Reference:

https://www.digitalocean.com/community/tutorials/how-to-install-anaconda-on-ubuntu-18-04-quickstart

[TUITORIAL] Install Scipy

sudo apt-get install python3-scipy

[NOTE] nvcc not found - darknet

which nvcc

->  /usr/local/cuda/bin/nvcc

export PATH=${PATH}:/usr/local/cuda/bin/nvcc

===========

[TUITORIAL] Build DeepStream SDK

sudo apt install \
    libssl1.0.0 \
    libgstreamer1.0-0 \
    gstreamer1.0-tools \
    gstreamer1.0-plugins-good \
    gstreamer1.0-plugins-bad \
    gstreamer1.0-plugins-ugly \
    gstreamer1.0-libav \
    libgstrtspserver-1.0-0 \
    libjansson4=2.11-1

sudo apt-get install librdkafka1=0.11.3-1build1
tar -xpvf deepstream_sdk_v4.0.2_jetson.tbz2
cd deepstream_sdk_v4.0_jetson
sudo tar -xvpf binaries.tbz2 -C /
sudo ./install.sh
sudo nvpmodel -m 0
sudo jetson_clocks
deepstream-app -c ./samples/configs/deepstream-app/source8_1080p_dec_infer-resnet_tracker_tiled_display_fp16_nano.txt

[ ] Reference:

https://docs.nvidia.com/metropolis/deepstream/4.0/dev-guide/index.html

[TUITORIAL] Install PyQt5 for jetson nano

Install PyQt5

sudo apt-get install python3-pyqt5

Install QtCreator

sudo apt-get install qtcreator

[TUITORIAL] Torch model 2 ONNX

import torch
import torchvision

MODEL_FACE_RECOGNIZE_PATH = 'BEST_checkpoint_r101.tar'
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
checkpoint = torch.load(MODEL_FACE_RECOGNIZE_PATH)
faceReg = checkpoint['model'].module
faceReg = faceReg.to(device)
faceReg.eval()

dummy_input = torch.randn(2, 3, 112, 112, device='cuda')

input_names = [ "actual_input_1" ] + [ "learned_%d" % i for i in range(16) ]
output_names = [ "output" ]

torch.onnx.export(faceReg, dummy_input, "insightfaceV2.onnx", verbose=True, input_names=input_names, output_names=output_names)

[TUITORIAL] ONNX to TensorRT

import os
import tensorrt as trt

batch_size = 2
TRT_LOGGER = trt.Logger()

def build_engine_onnx(model_file):
    with trt.Builder(TRT_LOGGER) as builder, builder.create_network() as network, trt.OnnxParser(network, TRT_LOGGER) as parser:
        builder.max_workspace_size = 1 << 30
        builder.max_batch_size = batch_size
        last_layer = network.get_layer(network.num_layers - 1)
        if not last_layer.get_output(0):
            network.mark_output(last_layer.get_output(0))
        with open(model_file, 'rb') as model:
            parser.parse(model.read())
        return builder.build_cuda_engine(network)

if __name__ == '__main__':
    onnx_file_path = './insightfaceV2.onnx'
    engine = build_engine_onnx(onnx_file_path)
    engine_file_path = './insightfaceV2.engine'
    with open(engine_file_path, "wb") as f:
        f.write(engine.serialize())

[TUITORIAL] .pb to TensorRT

[ ] Reference:

https://iwannatechcom.sharepoint.com/:w:/r/sites/iwanna-face/_layouts/15/Doc.aspx?sourcedoc=%7BA5AA2F4C-B3B8-4854-98CA-0913AF8227C1%7D&file=Speed%20up%20TensorFlow%20Inference%20on%20GPUs%20with%20TensorRT.doc&action=default&mobileredirect=true

code

from tensorflow.python import ops
import tensorflow.contrib.tensorrt as trt
import tensorflow as tf
import argparse

ap = argparse.ArgumentParser()
ap.add_argument("-m", "--model", type=str, help="path to pb file")
args = ap.parse_args()

def get_graph_def_from_file(graph_filepath):
    tf.compat.v1.reset_default_graph()
    with ops.Graph().as_default():
        with tf.io.gfile.GFile(graph_filepath, 'rb') as f:
            graph_def = tf.compat.v1.GraphDef()
            graph_def.ParseFromString(f.read())
        return graph_def

frozen_graph = get_graph_def_from_file(args.model)

output_names = ["boxes", "scores", "num_boxes"]
trt_graph = trt.create_inference_graph(input_graph_def=frozen_graph,
                                        outputs=output_names,
                                        max_batch_size=1,
                                        max_workspace_size_bytes=1 << 25,
                                        precision_mode='FP16',
                                        minimum_segment_size=3)

with open(args.model.replace(".pb", ".trt"), 'wb') as f:
    f.write(trt_graph.SerializeToString())

load TensorRT model

from tensorflow.python import ops
import tensorflow as tf
import argparse
import cv2

ap = argparse.ArgumentParser()
ap.add_argument("-m", "--model", type=str, help="path to pb file")
ap.add_argument("-i", "--image", type=str, help="image path")
args = ap.parse_args()

def get_graph_def_from_file(graph_filepath):
    tf.compat.v1.reset_default_graph()
    with ops.Graph().as_default():
        with tf.io.gfile.GFile(graph_filepath, 'rb') as f:
            graph_def = tf.compat.v1.GraphDef()
            graph_def.ParseFromString(f.read())
        return graph_def

if __name__=="__main__":
    graph_def = get_graph_def_from_file(args.model)
    graph = tf.Graph()
    with graph.as_default():
        tf.import_graph_def(graph_def, name='import')


    input_image = graph.get_tensor_by_name('import/image_tensor:0')
    output_ops = [
        graph.get_tensor_by_name('import/boxes:0'),
        graph.get_tensor_by_name('import/scores:0'),
        graph.get_tensor_by_name('import/num_boxes:0'),
    ]

    gpu_options = tf.compat.v1.GPUOptions(
        per_process_gpu_memory_fraction=0.5,
        visible_device_list="0"
    )
    config_proto = tf.compat.v1.ConfigProto(gpu_options=gpu_options, log_device_placement=False)
    sess = tf.compat.v1.Session(graph=graph, config=config_proto)

    img = cv2.imread(args.image)
    img_batch = img[None]
    boxes, scores, num_boxes = sess.run(output_ops, feed_dict={input_image: img_batch})
    
    to_keep = scores > score_threshold
    boxes = boxes[to_keep]
    scores = scores[to_keep]

    print(boxes, scores)

[ERROR] Extremely long model loading time problem of TF-TRT

Reference

https://jkjung-avt.github.io/tf-trt-revisited/

Reason

The default `python implementation` of python3 `protobuf` module runs too slowly on the Jetson platforms.

Solution

Replace it with `cpp implementaion` of that same module.

Install protobuf-3.6.1

wget https://raw.githubusercontent.com/jkjung-avt/jetson_nano/master/install_protobuf-3.6.1.sh
./install_protobuf-3.6.1.sh

[TUITORIAL] Object Detection in 10 lines code

sudo apt-get update
sudo apt-get install git cmake libpython3-dev python3-numpy
git clone --recursive http://github.com/dusty-nv/jetson-inference
cd jetson-inference/
mkdir build
cd build
cmake ../
make
sudo make install
sudo ldconfig

[NOTE] RTSP link

https://www.unifore.net/ip-video-surveillance/network-camera-rtsp-url-address-and-port-number.html

[RTSP] rtsp://admin:abcd1234@192.168.1.8:554/h265/ch1/main/av_stream

[Configure & Preview] http://192.168.1.8/doc/page/login.asp

[SADP Tool] https://www.hikvision.com/en/support/tools/destop-tools/sadp-for-windows/

[NOTE] GSTREAMER vs OpenCV

[v] - stream video, use NVDEC

gst-launch-1.0 filesrc location=~/Documents/your_name.mp4 ! qtdemux name=demux demux.video_0 ! queue ! h264parse ! omxh264dec ! nveglglessink -e

cap = cv2.VideoCapture("filesrc location=~/Documents/your_name.mp4 ! qtdemux name=demux demux.video_0 ! queue ! h264parse ! omxh264dec ! nveglglessink -e", cv2.CAP_GSTREAMER)

[v] - stream RTSP, use NVDEC

gst-launch-1.0 rtspsrc location=rtsp://admin:abcd1234@192.168.1.8:554/h265/ch1/main/av_stream ! decodebin ! nvvidconv ! video/x-raw,format=I420,width=1920,height=1080 ! videoconvert ! nveglglessink -e

cap = cv2.VideoCapture("rtspsrc location=rtsp://admin:abcd1234@192.168.1.8:554/h265/ch1/main/av_stream ! decodebin ! nvvidconv ! video/x-raw, format=I420, width=1920, height=1080 ! videoconvert ! appsink", cv2.CAP_GSTREAMER)

===========

[REFERENCE]

Use NVIDIA DeepStream to Accelerate H.264 Video Stream Decoding

[ ] https://medium.com/@fanzongshaoxing/use-nvidia-deepstream-to-accelerate-h-264-video-stream-decoding-8f0fec764778
[ ] https://docs.nvidia.com/metropolis/deepstream/4.0/dev-guide/
[ ] https://medium.com/tensorflow/high-performance-inference-with-tensorrt-integration-c4d78795fbfe

NVIDIA Jetson Nano courses

[ ] https://courses.nvidia.com/courses/course-v1:DLI+C-IV-02+V1/courseware/b2e02e999d9247eb8e33e893ca052206/5a688093f0504764beca665a33c2de7d/?child=last
[ ] https://gstreamer.freedesktop.org/documentation/application-development/introduction/gstreamer.html?gi-language=c

Failed to load driver Swrast

[ ] https://devtalk.nvidia.com/default/topic/1042427/linux/gtx-1080-steam-broke-driver-libgl-error-failed-to-load-driver-swrast/

Configuration Groups

[ ] https://docs.nvidia.com/metropolis/deepstream/4.0/dev-guide/index.html#page/DeepStream_Development_Guide%2Fdeepstream_app_config.3.2.html%23

Keyword: Convert tensorflow model to .uff
[ ] https://github.com/smistad/convert-tensorflow-model-to-tensorrt-uff
[ ] https://github.com/NVIDIA-AI-IOT/tf_trt_models/
[ ] https://docs.nvidia.com/deeplearning/sdk/tensorrt-archived/tensorrt_401/tensorrt-api/python_api/index.html
[ ] https://devblogs.nvidia.com/accelerate-video-analytics-deepstream-2/

Deep Learning SDK Documentation

[ ] https://docs.nvidia.com/deeplearning/sdk/tensorrt-sample-support-guide/index.html

Gst.py

[ ] https://gitlab.freedesktop.org/gstreamer/gst-python/-/blob/master/gi/overrides/Gst.py

Note 02/03/2020

[ ] https://devtalk.nvidia.com/default/topic/1049399/tensorflow-model-to-tensorrf-uff-model-convert-error/
[ ] https://github.com/NVIDIA-AI-IOT/tf_trt_models/blob/master/examples/detection/detection.ipynb
[ ] https://docs.nvidia.com/deeplearning/sdk/tensorrt-archived/tensorrt-601/tensorrt-api/python_api/parsers/Uff/pyUff.html

Note 03/03/2020

[ ] https://github.com/IntelAI/models/blob/master/benchmarks/object_detection/tensorflow/ssd-mobilenet/README.md#int8-inference-instructions
[ ] https://github.com/tensorflow/models/issues/5300

SSD model

[ ] https://github.com/IntelAI/models/blob/master/benchmarks/object_detection/tensorflow/ssd-mobilenet/README.md#int8-inference-instructions

Note 09/03/2020

[ ] https://blog.tensorflow.org/2018/04/speed-up-tensorflow-inference-on-gpus-tensorRT.html

RetinaNet to TF Serving,Freezing and Optimisation
[ ] https://colab.research.google.com/drive/1u79vDN4MZuq6gYIOkPmWsbghjunbDq6m#scrollTo=CwxD2I2U7zbu&forceEdit=true&sandboxMode=true


frozen_file = "saved_model.pb"
frozen_file = "ssdmobilenet_int8_pretrained_model.pb"
output_nodes = ["detection_scores", "detection_boxes", "num_detections", "detection_classes"]
uff.from_tensorflow_frozen_model(frozen_file, output_nodes, output_filename="output.uff")


DIR=${DIR:=`pwd`}

===========

[TUITORIAL] Encrypt python code

[ ] Reference: Step 1:

Change
    `if __name__ == '__main__':`
into
    `def main():`
in file containing main function.

Step 2:

python3 encrypt.py -s /source -e js_gui.py:main -o /source/build

eg:
    python3 encrypt.py -s /home/jetson/Documents/ -e js_gui.py:main -o /home/jetson/Documents/build

[TUITORIAL] request

curl -X POST http://x.x.x.x:8080/predictons/getembed -T "m.jpg"

[TUITORIAL] CRONJOB

Create job

#pip install python-crontab
from crontab import CronTab

my_cron = CronTab(user=True)
job = my_cron.new(command='~/Downloads/ngrok tcp --region=ap 22', comment="ngrok")
job.every_reboot()
my_cron.write()

my_cron = CronTab(user=True)
job = my_cron.new(command='cd ~/Documents/face_stream; python3 manage.py runserver 0.0.0.0:8003', comment="runserver")
job.every_reboot()
my_cron.write()

Remove job

from crontab import CronTab
my_cron = CronTab(user=True)
for job in my_cron:
    if job.comment == 'ngrok':
        my_cron.remove(job)
        my_cron.write()
        print('Removed ngrok')
    elif job.comment == 'runserver':
        my_cron.remove(job)
        my_cron.write()
        print('Removed runserver')
    else:
        print(job.comment)