My Pi 3 Install Log (6): OpenCV (part 2) on Raspberry Jessie (Lite)

(4 hours)

# in part 1 after creating a virtual env for python 2.7, i changed my mind.

# thought like redo in python3 because my physical env is python2 and has numpy installed already.

$ deactivate # leave env

$ rm -rf .virtualenvs/cv

$ mkvirtualenv cv -p python3

Already in, have (cv) before prompt

If not in (cv) yet

$ source .profile

$ workon cv

pip list # show what has been installed

pip install numpy # took quite long time ~10mins!!

# version 1.11.1

$ cd ~/opencv-3.1.0/

$ mkdir build

$ cd build

$ cmake -D CMAKE_BUILD_TYPE=RELEASE \

   -D CMAKE_INSTALL_PREFIX=/usr/local \

   -D INSTALL_PYTHON_EXAMPLES=ON \

   -D OPENCV_EXTRA_MODULES_PATH=~/opencv_contrib-3.1.0/modules \

   -D BUILD_EXAMPLES=ON ..

# I want to see cmake output so startx and ran from lxterminal.

# it was a mistake to also call ‘make’ there took me 102mins while the instruction said 75mins.

$ make -j4

# but it completed successfully.  :) 2.7GB was generated within the build folder

# install to system

$ make install

$ ldconfig

## step 6

$ cd /usr/local/lib/python3.4/site-packages/

$ ls -l

total 1852

-rw-r--r-- 1 root staff 1896140 Aug 22 21:43 cv2.cpython-34m.so

$ sudo mv cv2.cpython-34m.so cv2.so

$ cd ~/.virtualenvs/cv/lib/python3.4/site-packages/

$ ln -s /usr/local/lib/python3.4/site-packages/cv2.so cv2.so

* * *
congrats, installation completed

* * *

## step 7 verify installation

# open a new terminal

$ source ~/.profile

$ workon cv

$ python

>>> import cv2

>>> cv2.__version__

'3.1.0'

>>>

## my step 8 (optional)

The two opencv folders at ~ no longer necessary for python code.  
Before moving on, I want to back up the 'build' folder for what have been compiled and generated during 102 mins wait.

# plug in a big enough USB disk, say have a label myusbdisk

$ tar cvf /media/pi/myusbdrive/opencv-3.1.1_build.tar /home/pi/opencv-3.1.0/build/

(edited)
built samples and examples in ~/opencv-3.1.1/build/bin
to try them, first start the gui
open a terminal, and enter
sudo modprobe bcm2835-v4l2 # see ref below

sometimes example stuck won't quit
kill that process
1. ctrl-shift-n for a new terminal
2. ps ax | grep facedetect
3. (note the pid, eg. 12345)
4. kill 12345


* * *

ref:

https://www.raspberrypi.org/forums/viewtopic.php?f=43&t=79731

I think the difference is that he is not using the Raspberry Pi camera but a USB webcam - at 4:40 guvcview is reporting using driver uvcvideo. I've just tried it and guvcvideo fails immediately for me complaining about the contents of /proc/cpuinfo (no CPU frequency, although I don't know why it needs to know that!). 

I haven't listened through all of the tutorial, but I think the bit you need is 
CODE:  sudo modprobe bcm2835-v4l2
to load the RaspiCam V4L2 driver. raspistill and raspivid talk to the MMAL service directly, whereas most applications (including OpenCV) use V4L2.

* * *

# to build cpp examples myself   https://www.learnopencv.com/how-to-compile-opencv-sample-code/

# Gcc using opencv

You can use pkg-config command to get the compiler/linker options. Like this

gcc `pkg-config --cflags opencv` `pkg-config --libs opencv` new.c
Please note the above punctuation is ` not '.

Or you can use CMake. Write a CMakeLists.txt like this

cmake_minimum_required (VERSION 2.6)
project (new1)
find_package (OpenCV REQUIRED)
include_directories (${OpenCV_INCLUDE_DIRS})
add_executable (new1 new.c)
target_link_libraries (new1 ${OpenCV_LIBS})
then you can build like this

mkdir build
cd build
cmake -DOpenCV_DIR=<the-directory-contating-OpenCVConfig.cmake> ..
make