LaneNet Install in NVIDIA Jetson Xavier AGX

===================2020.10.24================================

See 2020.12.14 update

 

Download or git clone LaneNet :

$ git clone github.com/MaybeShewill-CV/lanenet-lane-detection.git

github >> github.com/MaybeShewill-CV/lanenet-lane-detection

MaybeShewill-CV/lanenet-lane-detection

you may install them by

$ pip3 install -r requirements.txt

 

Execution command:

$ python3 tools/test_lanenet.py --weights_path /YOUR/CKPT/PATH/tusimple_lanenet.ckpt --image_path ./data/tusimple_test_image/0.jpg

* Requirements Error I encountered & Version Check

$ pip3 install -r requirements.txt

When i  install requirements, I get this error.

ERROR: No matching distribution found for opencv (from -r requirements.txt (line 7))

 

But, I ignored this.

 

I don't konw if the order of installation is important.

First, install it

$ sudo apt-get install python3-sklearn python3-sklearn-lib

 

1. numpy

Set numpy version to 1.18.1

It has a version conflict issue with scipy, so be sure to set it to 1.18.1

 

check own version :

$ pip3 show numpy -V

 

Then it comes out like it :

Name: numpy

Version: 1.18.1

 

So, if you already have numpy, I recommend installing it according to the version I tried.

$ pip3 install numpy==1.18.1

 

2. glog

Set glog version to 0.3.1

 

I remember installing glog several times.

I can't remember how i installed it, but I installed the latest version.

 

3. tensorflow_gpu

$ pip3 install tensorflow-gpu

 

Then, check tensorflow_gpu in python3

 

$ python3

>>> import tensorflow as tf

>>> hello = tf.constant('Hello!')

>>> sess = tf.Session()

>>> print(sess.run(hello))

b'Hello!'

 

4. scikit_learn

This made me troublesome!

I did a lot of installation and somehow it worked out.

I will tell you the version of scikit_learn and scipy that i have installed, so please install it once.

 

$ pip3 show scikit_learn

Name: scikit-learn

Version: 0.19.1

 

$ pip3 show scipy

Name: scipy

Version: 1.5.2

 

I did this too.

$ pip3 install scipy --upgrade

 

 

*My opencv_version is 3.3.1 :

$ opencv_version

3.3.1

 

Good Luck!!!

 

===================2020.12.14================================

I try installing to lanenet on other agx xavier

so, i update this installation

(my opencv version is 3.3.1 as default)

 

First, same above

$ git clone github.com/MaybeShewill-CV/lanenet-lane-detection.git

 

Go lanenet-lane-detection dir and vim requirements.txt

 $ vim requirements.txt  // change it like this - remove opencv, tensorflow-gpu

$ pip3 install -r requirements.txt

 

After pip3 installation, i install tensorflow_gpu

$ sudo apt-get install zlib1g-dev zip libjpeg8-dev libhdf5-dev

$ sudo pip3 install -U numpy grpcio absl-py py-cpuinfo psutil portpicker grpcio six mock requests gast astor termcolor
$ pip3 install --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v411 tensorflow-gpu

 

Check tensorflow_gpu installation in xavier

Python 3.6.7 (default, Oct 22 2018, 11:32:17)

[GCC 8.2.0] on linux

Type "help", "copyright", "credits" or "license" for more information. 

>>> import tensorflow 

>>> import tensorflow as tf 

>>> tf.__version__ 

'1.12.0'

출처: https://goodtogreate.tistory.com/entry/Jetson-AGX-Xavier-동작-모드-변경-및-TensorFlowGPU-설치와-실행-그리고-성능-분석 [GOOD to GREAT]

 

Done? so, you can run lanenet

$ python3 tools/test_lanenet.py --weights_path model/tusimple_lanenet.ckpt --image_path ./data/tusimple_test_image/0.jpg

 

After you run lanenet, if you have error, See above.