Open source vision page

###### Software installation guide####################################
Check versions #2023 Feb 15
>conda -V #inside conda, check conda version, e.g., conda 22.9.0
>python #go inside python , its version will be shown, .e.g pythion 3.9.13
   >>> import cv2 #start opencv
   >>> cv2.__version__ #will show opencv version. E.g. python-opencv-'4.7.0'
>>> control_Z to quit python

#######################################################
### Important notes, 2023 Feb 15    #

##############################################################
Public domain software.pptx

inside cuhk you need to use --proxy for installation, i.e., to install graphviz

pip install --proxy http://proxy.cse.cuhk.edu.hk:8000 graphviz

# if you work inside CUHK, you should specify proxy in the file ".condarc" .
ssl_verify: false
channels:
- defaults
proxy_servers:
http: http://proxy.cse.cuhk.edu.hk:8000
https: http://proxy.cse.cuhk.edu.hk:8000

###############################################################
### Installation and testing of colab (2023 Feb. 7) ########
###############################################################

1.    Open a Gmail account
2.    You should be provided with a google drive.
3.    There are plenty of tutorials, such as https://www.youtube.com/watch?v=RLYoEyIHL6A
4.    Now you can start to try some tutorials of colab. Such as
    a.    https://colab.research.google.com/#scrollTo=-gE-Ez1qtyIA
    b.    You may choose tutorial such as MNIST.ipynb in https://colab.research.google.com/github/skorch-dev/skorch/blob/master/notebooks/MNIST.ipynb

Using colab to try the demo of decision tree in sklearn
https://colab.research.google.com/github/jackqk/sklearn-note/blob/master/Decision_Tree.ipynb

###############################################################
### Installation anaconda (2023 Jan. 10) and opencv ########
###############################################################

Install anaconda first: https://www.anaconda.com/products/distribution, https://www.datacamp.com/tutorial/installing-anaconda-windows

Fix: https://stackoverflow.com/questions/71204741/how-to-fix-error-module-cv2-has-no-attribute-legacy-on-python-3-7-9-and-w
==== python version problem, if you need python 3.7 (or some other versions) for your application, you may change it inside "anaconda power prompt" do

To use aruco tracker in https://github.com/njanirudh/Aruco_tracker , use python 3.7.13, opencv-contrib-python==3.4.5.20 (tested ok)

>conda install python=3.7.13
> pip uninstall opencv-contrib-python #remove other version first.

> pip install opencv-contrib-python==3.4.5.20
///////////////
###############################################################
### Setup Virtual environment under conda 2023. Feb 15    ########
###############################################################

>conda -V                  # Step 1: ?conda is installed,version

>conda update conda        #Step 2: Update conda
>conda search “^python$”   #Step 3: check possible python ver.
>conda create -n envname python=x.x anaconda #Setup vir.Env.
>conda activate your_envname #Step 4: Activate virtual envir.

# now in your envi. called your_envname
>conda install -n your_envname package #install pacakges

#if you are in cuhk, you need to specify proxy
>conda install --proxy http://proxy.cse.cuhk.edu.hk:8000 -n your_envname package

>conda deactivate #Step 6: Deactivate. your_envnem and quit virt.

#IF you need this environment anymore
>conda remove -n envname -all

###############################################################
### Setup opencv microsoft visual studio,cpp python 2023, Jan 10 ######
###############################################################

vs2017-opencv-Installation.docx ,

vs2019-opencv-Installation.docx ,(2023 Jan 10)

opencv4_sample_run_result.docx

###############################################################
### Others ######
###############################################################
Python tricks:
>autopep8 fact.py --in-place #auto fix python file format, fix space tab problems
>2to3 -w frame.py # fix python 2 programs to python3

On Visual studio:sing Visual Studio Professional 2019 (version 16.11) https://my.visualstudio.com/Downloads?q=Visual%20Studio%202019

Source: https://docs.microsoft.com/en-us/visualstudio/releases/2019/history

To enter arguments to a program under visual studio IDE

Download the lk_track , camshift sample code from OpenCV Github repository

( https://github.com/opencv/opencv ) (C++ or Python), After unzip , python samples codes are in \opencv-4.x\samples\python

Demos: run lk_track.py, lk_track_save_data.py (http://www.cse.cuhk.edu.hk/~khwong/www2/cmsc5711/lk_track_save_data.py) and camshift.py

§ =========== Updated 2022.3.12, if you have problem using opencv (4.4.x.x etc), try to install an older version, follow the following methods

§ conda> pip uninstall opencv-python

§ conda> pip uninstall opencv-contrib-python

§ conda > pip install opencv-python==3.4.2.17 ,

§ conda > pip install opencv-contrib-python==3.4.2.17

§ =====

§ You may try the following samples from

§ https://github.com/opencv/opencv/tree/4.x/ (download the zip file and find the sample codes)

§ conda> python camshift.py 0 # camshift using camera 0 , use mouse to select area to track

§ conda> python lk_track.py 0 # lk tracker for using camera 0 (laptop camera)

§ conda> python calibrate.py #using input image files from ..\data, you my use your own ima

///////////////////////////////////////////////////////////////////////////////////////////////////////////

/////////// Opengl installation under anaconda and python if you have problems of using opengl in python////
Install opengl using anaconda_python.
Step1: Check your python version, assume you are in “D:\_project223\teapot_py”
(base) PS D:\_project223\teapot_py >>python –version
e..g you get : Python 3.7.11
////////////////////////////////////////////////////////////////////////////////////////////////////////
step2a: check your python version:
(base) PS D:\_project223\teapot_py> python
Python 3.7.11 (default, Jul 27 2021, 09:42:29) [MSC v.1916 64 bit (AMD64)] :: Anaconda, Inc. on win32
Type "help", "copyright", "credits" or "license" for more information.
The above shows it is 64 bit (AMD).
////////////////////////////////////////////////////////////////////////////////////////////////////////
step2b: check tensorflow version if you have installed tensorfow
>>pip show tensorflow
or
>>pip list
//////////////////////////////////////////////
Step3 : download files, goto https://www.lfd.uci.edu/~gohlke/pythonlibs/#pyopengl
find the suitable 2 files for installation, since we use python 3.7.11/64 bit ,use cp37, and 64bit AMD
conda >> pip install PyOpenGL_accelerate-3.1.6-cp37-cp37m-win_amd64.whl
conda >> pip install PyOpenGL-3.1.6-cp37-cp37m-win_amd64.whl
If you are successfully then you are ok. If not you may get the wrong version of .whl, check again.

Demo programs
teapot_test.py
demo_opengl.zip

///////////////////////////////////////////////////////////////////////////////////////////////////////////

/////////// CPP (C++) (32-bit) Opengl installation
Unzip Glew and Glut.zip and place the whole directory in c:\Glew and Glut
In visual studio (e.g. vs2019 ) set the correct directories in C++,and link_dir of the project see the example (see .sln): opengl1_teapot.zip

//////////////////////// older ///////////////////////////////

---- Dec 2016/ 2017-------

Summer project webpage 2013 (summer)

Python+OpenCV on Linux by Bill yau

latest version of ba ba810c.zip

installation guide

---------2014--------------------------------------------------------------------

Open source Vision and Image Processing library (OSVIP)

Updated 25.11.2014

Download sites (share_vision), KK

Demos

Deep Net

Caffe_setup_guideV5a

Andriod+Opencv :

andriod_opencv_lkdemo_install_guide2014
http://opencv.itseez.com/doc/tutorials/introduction/android_binary_package/android_binary_package.html
Settings of the Android development enviroment ( Eclipse + Android SDK ), you may refer to this page: http://developer.android.com/sdk/installing.html

Install_python2.7_opencv2.4_win7_or_xp_vs2010.doc
Theory

Camera calibration used in Opencv (step-by-step-Zhang's method Web , Zhang99, Zhang Pami 2000 , ppt)

Current work:

Python 2013

readme_python27_opencv472b.doc

to be done

see http://www.t-immersion.com/ to seeif we can produce a similar tool.
add video read/write to Nvision

Nvsion and TabBox projects are written in c# using vs2010, they also use some functions of OpenTK and EMGU.
C# in Visual Studio vs2010 tutorials

OpenTK

http://www.opentk.com/book/export/html/111 (a book)
http://www.opentk.com/doc //show you how to write a simple OpenTK program in 15 mins.

EMGU (wrapper for opencv to be used with C#,C++ etc)

Installation,

http://www.emgu.com/wiki/index.php/Main_Page
Tutorial Code in EmguCV which is : http://www.emgu.com/wiki/index.php/Code_Gallery

Tutorials

http://fewtutorials.bravesites.com/tutorials

Tutorial for beginners:http://fewtutorials.bravesites.com/entries/emgu-cv-c/level-1---lets-make-a-camera-application

OpenTK which is an OpenGL library in C# syntex: http://www.opentk.com/

OpenGL

3D models

http://www.cc.gatech.edu/projects/large_models/
http://mview.sourceforge.net/ mesh viewer

opencv2.4 documentation, Opencv2.4_tutorials

Installation : http://www.cnblogs.com/freedomshe/archive/2012/04/25/2470540.html,

install_python2.7_opencv2.2_or2.3_win7_or_xp_vs2008.doc (update 8.5.2012) lib230d.txt

opencv2.3 with python2.7 tested ok (opencv2.3 provides more python functions though cv.pyd and cv2.pyd)
FAQ_opencv_vs2008.doc
Python_examples (tested ok programs)

Demo vision:

Recent Quadrangle detection 20Aug2012 :

Our quadrangle detection demos, 39 July 2012. The demo videos show the newly proposed quadrangle detection method.
1. First it detects Q Corner keypoints from the hough lines,
2. Then the quadrangle is determined from the Q Corners.
3. The false positive quadrangle observations are not filtered in the demo videos.
This newly proposed quadrangle detector will be used in the calibration toolbox for detecting the cardboard. Therefore no more manual corner selection is required (old programs are required manual corner selection). The new method can have multiple quadrangle observations instead of only one observation in the old applications. The new method does not depends on overlapping ratio of the quadrangle edge.

3 July 2012 ACCV2012 demo http://sdrv.ms/MPLVOO
22 June . particle filter face tracking:
http://www.youtube.com/watch?v=jFqiw5y8Qp0&feature=youtu.be
20 June 2012

The updated video clips of face tracking:
[eye coordinate is fixed: w.r.t. camera coordinate]
(cube replacing face)
http://www.youtube.com/watch?v=TspKqJfEVTA&feature=youtu.be
(teapot replacing face)
http://www.youtube.com/watch?v=4XhCdylIowE&feature=youtu.be
The above videos are with 4*4 M_view = estimated 3D pose, and 4*4 M_model = 150 units z translation, where M_view * background and M_view * M_model * object
[eye coordinate is changed: w.r.t. world coordinate]
http://www.youtube.com/watch?v=zPHIA2_d0rE&feature=youtu.be