Kartik's blog: January 2016

Friday, 29 January 2016

Which Linux Distro to choose as a begineer ?

As mentioned in the earlier post, Linux has become an integral part of our life so now one cannot simply escape without knowing/using Linux. We may see in various hacker movies, the operating system look as though there is only text (Users with a Linux disto simply press ctrl+alt+f1 to know) and we need to know some sophisticated language and type with incredible speed etc.. THIS IS NOT THE CASE AT ALL. Everyone has a first time with Linux so it is not necessary to know commands for the terminal while using Linux.

Most of the Linux distros come with a desktop environment so that beginners do not face any issue when starting. Only on some rare occasions you may have to use the terminal to write commands e.g. installing libraries for a C/C++ program, building a new software from scratch, etc... There are many different Desktop Environments like and my preference order is as follows:

Cinnamon
Mate
GNOME
Unity (Mainly used by Ubuntu)
KDE
XFCE
LXDE
Standard i.e. (no desktop environment so only commands like Debian standard)

For a beginner, it would be nice to use a distribution which is easy to install as well as use. So the best way to start is to use "Ubuntu" (as it has support for UEFI ) or "Linux Mint" (Linux Mint 17 in case you are using an older system with BIOS and 17.3 "Rosa" support UEFI) or "Elementry OS"(which gives the feel of Mac OSX using Pantheon desktop) "Fedora" (a bit tedious while installing) which is also UEFI compatible. If you are completely new to Linux, the standard Ubuntu with Unity Desktop will be most preferred.

Ubuntu or its derivatives are easy to install as they generally have the "install alongside" option. One Distro (say Ubuntu with default Unity Desktop) can come in different other desktop environments , as mentioned above, so simply have a look at the other distros (Kubuntu, Lubuntu, Ubuntu-Mate, Ubuntu Studio, etc.. ) from distrowatch.com and see which one appeals the most. So have a look at the distro's desktop
(search for screen shots of the distro you are interested in).

My personal preference is the cinnamon desktop which is not present in Ubuntu. Linux Mint is a derivative of Ubuntu so all the normal command work with a few changes e.g. bashrc file. Next i would choose Ubuntu Mate as it preserves the look of Ubuntu 10.04 if one needs to know how to use / locate applications in the environment. It is not as resource hungry as the later versions of Ubuntu (>13.04) with unity desktop. But then again you are welcome to explore all the distros.

Those who are well versed with Linux can try distributions like Debian which is a very stable and sturdy distro or Fedora which is a project of Redhat. or even Arch Linux, Slackware,

Operating Systems

Linux has now become a part of our life thanks to Linus Torvalds. Many of our systems use Linux in the backend and we are generally unaware e.g. Xbox, Playstation, Web Servers, Cloud computing servers (e.g. Amazon), Android smartphones (Android OS is a linux), Backtrack/ Kali (the famous hacking OS's) etc.. So it is suggested to try atleast one of these flavors of linux from the wide variety available from the net.

As many of you have seen there has been a huge change in the computers we had in the 90's and in recent times. Moore's law is holding true so far that the number of transistors are doubling. Because of this fact our latest machines are capable to perform many tasks and give us a faster and better way to perform the tasks. Many processors came and came with them the operating system, our guides to the hardware. Generally there are 3 types of operating systems known to most users :

Windows
Mac OS
Linux

Each have their pros and cons like Mac OS was designed to optimized performance and battery life same as that of Windows. It was battle between these two titans, until the arrival of Linux. I have used all the above operating systems and concluded the following:

Anybody can use a Mac and it is good in terms of battery life. Windows however, the User Interface (UI) keeps changing with each version(windows XP, 7, 8, 10) so may be slightly hard to use and Battery life does not last more than three hours. Linux may be slightly tough for a beginner but once you get used to it you start to appreciate it however no improvement in battery life.
Mac and Windows are proprietary OSs so one cannot just download the OS for free (legally). Linux on the other hand is supported by different organizations and is free for anyone to use as a server, as a desktop, as a cloud, etc.. So many organizations prefer to use Linux on their servers and networks
"No one size fits all" .... So Linux comes in many different distributions like Debian, Ubuntu, Linux Mint, Backtrack, Elementry, etc.. are all different flavors of the Debian OS with slight tweaks depending on the way you like it.
After Windows 8.1, the normal BIOS has been changed by UEFI which makes it slightly difficult to install another OS along side Windows 8.1 and above. You can only install those Linux OSs which do have support for UEFI like Ubuntu, Fedora, etc..
There is no way you can try out Windows or Mac OS without installing them. Linux can be written into a CD / USB flash drive to make "Live" CD/ USB so that one can simply try out the OS before installing
Windows is very easily susceptible to virus attacks so an anti-virus program is always required. Mac may not need anti-virus programs as it partially resembles a version of Linux. Linux does not need anti-virus programs.

Running openCV

Running Programs in openCV:

There are many ways to run .c or .cpp programs in opencv you can use the methods specified in the homepage.
I am following the method mentioned in the opencv homepage for using gcc and eclipse

Using Cmake:

To run opencv programs one must now:

Create a project folderi.e.,say a program 'proj_name.cpp' is to be created compiled and executed. Create a Folder proj_name (This is only for convinence sake). Create the source file proj_name.cpp

Create a .txt file named "CMakeLists.txt" and it must contain the following

cmake_minimum_required(VERSION 2.8)

project( proj_name )

find_package( OpenCV REQUIRED )

add_executable( proj_name proj_name.cpp )

target_link_libraries( proj_name ${OpenCV_LIBS} )

Now to compile and create an executable object use the following commands:

cd <DisplayImage_directory>

cmake .

make

4)After compiling you will get an executeable object. To run the code use the command

./proj_name any_parameter_needed

any_parameter_needed varies depending on your program

Using eclipse(C/C++):
I am following the method mentioned in the opencv homepage

openCV Sample tutorials

To understand the syntax of using openCV in C/C++ it is better "Follow by example". A small tutorial has been generated to understand how to implement the following operations in Matlab/Octave and openCV in C++ :

Mapping images into different colour spaces:
This part of the program deals with reading an image, understanding its data-structure, seperating different planes (in case of colour images) , conversion from RGB (Red,Blue,Green) to HSV (Hue,Saturation,Value) plane, YUV plane(Intensity, Complement of red, Complement of blue plane)
Point Processing:
This section deals with how different point processing techniques (Where region of the pixel does not matter) such as Negative of an image, Gray level slicing, Contrast Enhancement, Bit plane slicing etc..
Neighbourhood Processing:
This section deals how to use masks and convolution operators for Low pass filters, High pass filters, etc..

openCV App: Vehicle Detection and counting

This is a simple program implemented in opencv to detect and count the number of vehicles passing through a virtual fence. A virtual fence is an imaginary line/region of interest (ROI) which is used in detection and processing of a moving object. A virtual fence can be useful for counting of an object, detection of intrusion, determining direction of a moving object but for this purpose we have limited the function of our virtual fence to detection and counting. However, many a priori assumptions were made for determining the optimality taking into account environmental factors such as location and positioning of the camera, illumination variation, ambient noise etc..

The procedure is as follows

An inital frame is kept as a reference (prev_frame) and later the next frame (current_frame) is taken and the forward difference(current_frame - prev_frame ) is calculated.
Later this image is thresholded to a value such that the noise is suppressed and the objects and highlighted and median filtering is done to remove any noise which might have come after thresholding.
After thresholding, we see that the object is not connected properly so morphological operators are used to join the object.

https://youtu.be/gPC1y1FpQwc

openCV comaptibility with other languages

Matlab is good when it comes to image processing but the codes used are also written in the same language hence the execution of the code lags. Here is where openCV has the advantage as the codes are written to be fast. It is now possible to use openCV in Matlab as well. However I have not found any good document which works perfectly.
OpenCV is mainly developed in C /C++ and they have been optimized in these languages however C / C++ can become complex at times so some developers use python to integrate with openCV and reuse the C/C++ code. OpenCV has python bindings which allows python use the C/C++ functions in the library.This is a good feature to those who are not so familiar with coding (Python is very easy to learn) and for using additional functionality e.g. python can handle complex numbers directly while C/C++ does not deal with it directly.

openCV installation

Installation in Windows:

On the home page of opencv,they have already mentioned how to install the openCV library in windows and configuring it to visual studio.
It is not necessary that you need visual studio. Any C++ compiler/ IDE can be used such as codeblocks compiler (Courtesy STAB IIT Bombay check the link).
First Download the codeblocks and select "codeblocks-13.12mingw-setup.exe" file. After the installation is complete reboot the system.
Download the opencv2.0 library for windows and during the installation you will be asked about the system path.
Select the option "Add openCV to the system path for all users".
Later in the compiler options-->Search directories-->Compiler and add the include folder of openCV. (Note: BE SURE NOT TO KEEP THE PATH RELATIVE FOR THESE STEPS.)
Next goto compiler options-->Search directories-->Linker and add the lib folder

Next goto compiler options-->Linker Settings and add all the contents of the lib folder one at a time.
Thats it. Now openCV has been configured and can be used to start implementing.
Note: openCV 2.0 has only the basic libraries (which do not include the folder opencv2) so for an introduction to image processing using openCV it is useful. If the advanced functions ( > 2.0) are required the please follow the steps for openCV>2.0 However the installation of this library requires additional 10 software(CMake,tortoise SVN/Git, Miktex, Python, Numpy, Intel Threading blocks, CUDA, Eigen, openEXR, openNI and SphinX) to to downloaded and installed separately. Thus it is recommended to use linux (Ubuntu, Fedora, Linux Mint, Debian etc..) for using openCV.
___________________________________________________________________________________________________________________________________________________

Installation in Linux:

If you check out the home page of opencv, they have already mentioned how to install the software in linux however, I faced some difficulties after the installation so I checked for errors and dependency issues and made a complete step-by-step procedure for installing openCV in Linux. You an also follow the steps mentioned in this web page if you are facing those issues.

Note:I have tested on Ubuntu and Linux Mint so far. It should work for other Debian based distributions as well.

The following are my sources if you would like to explore. Enjoy!!!

Installation Steps:

These instructions can be directly copy-pasted to the terminal in linux. The only assumption is that there is a working Internet connection. It supports the latest openCV currently version 3.1.0.

Install all pre requisites

Ubuntu provides a package of OpenCV 2.x.x that you can easily install by typing:

sudo apt-get install libopencv-dev

If you do not care about having the latest version you could skip the rest of the post however while interfacing with eclipse you may face some issues so it is recommended to follow all the instructions.

Depends on which version of linux you are using.

Simply Copy paste the following into your terminal screen it will install the necessary softwares and dependencies required.

sudo apt-get install libopencv-dev build-essential checkinstall cmake pkg-config yasm libtiff4-dev libjpeg-dev libjasper-dev libavcodec-dev libavformat-dev libswscale-dev libdc1394-22-dev libxine-dev libgstreamer0.10-dev libgstreamer-plugins-base0.10-dev libeigen3-dev libv4l-dev python-dev python-numpy libtbb-dev libqt4-dev libgtk2.0-dev libpng12-0 libpng12-dev libpng++-dev libpng3 libpnglite-dev zlib1g-dbg zlib1g zlib1g-dev libjasper-dev libjasper-runtime libjasper1 pngtools libtiff4-dev libtiff-tools libjpeg8 libjpeg8-dev libjpeg8-dbg libavcodec-dev libavformat-dev libgstreamer0.10-0-dbg libgstreamer0.10-0 libgstreamer0.10-dev libxine1-ffmpeg libxine-dev libxine1-bin libunicap2 libunicap2-dev libdc1394-22-dev libdc1394-22 libdc1394-utils swig libv4l-0 libv4l-dev libeigen2-dev libfaac-dev libopencore-amrnb-dev libopencore-amrwb-dev libtheora-dev libvorbis-dev libxvidcore-dev

(Even after here also you can use openCV however to ensure that the library is properly configured please proceed ahead)

Installation from the tar file

Download the source code from openCV homepage.Next go to the directory where OpenCV is downloaded via the terminal only and then un zip the package:

Note:- It is recommended that you move the downloaded OpenCV package to the home/<user>/

unzip OpenCV-2.x.x.zip

cd OpenCV-2.x.x/

Now make a new directory (folder) called "build" for building the software. Use the following commands.

mkdir build

cd build

Later run Cmake with the following commands

cmake -D WITH_TBB=ON -D WITH_QT=ON -D WITH_XINE=ON -D BUILD_NEW_PYTHON_SUPPORT=ON -D WITH_V4L=ON -D INSTALL_C_EXAMPLES=ON -D INSTALL_PYTHON_EXAMPLES=ON -D WITH_OPENGL=ON -D BUILD_EXAMPLES=ON ..

Now make the files

make

Now we make it permanent

sudo make install

Configuring OpenCV:

The installation is nearly complete so now we have to configure with the OS to use gcc or any ide(e.g. eclipse) to compile and run the programs.

sudo gedit /etc/ld.so.conf.d/opencv.conf

Add the following line at the end of the file (it may be an empty file, that is ok) and then save it:

/usr/local/lib

Close the file and run the following command to configure the library:

sudo ldconfig

OR we can alternatively run the command

export LD_LIBRARY_PATH=/usr/local/lib

Open your .bashrc file and add the following:

sudo gedit /etc/bash.bashrc

append the following at the end of the file -

PKG_CONFIG_PATH=$PKG_CONFIG_PATH:/usr/local/lib/pkgconfig

export PKG_CONFIG_PATH

Save and close the file and reboot the system

Installation is now complete.

openCV intro

Generally when we think of image processing we think of a famous software "Matlab". Matlab has made coding easy for novice beginners even outside the world of image processing but many have a notion that Matlab is the only tool to be used.Matlab is a high level language for researchers and students to develop things with ease.Traditionally if a task had to be done (say simply reading an image) the level of C/C++ programming required would have to be huge thus limiting the field to those who could write efficient program. This is where Matlab was useful as it removed the base level programming(creating pointers, allocate memory reading the data) required and focused on the abstract layer programming(Theories to be developed). Matlab has its perks if you can optimize your code in matlab to execute it in shorter run time.Also if you have the finance to purchase the product. But I want inform that there are other tools which can be used instead of Matlab such as Octave, openCV, etc..
OpenCV is a very powerful open source library for image and video processing applications and can be used in any platform such as Windows, Linux, Android, iOS,Raspberry pi, etc...You can check out their website opencv. I will be sharing my experiences and tricks with openCV. Lets start with installation of openCV in both Windows and Linux.I will be following the Mat structure instead of IplImage structure as it is a bit easier to program.

openMP

It is a pain that each and every library has to downloaded and installed and configured but it is a necessary evil in most cases. If we had each and every tool in our computer then we would have no space and our systems would be slow as there are other tools which we don't know about. For those who are using:

Windows

If an Intel compiler is already present in Windows, simply launch the software development environment (Intel parallel studio XE) and you will get a command prompt. Next type icl /QopenMP c_code.cpp.
The standard compiler for Windows comes from Microsoft Visual C/C++ compiler from Visual Studio. Unfortunately its OpenMP support is a bit outdated. To use it simply Start a project in Visual studio.

Start a new Project
Select a Win 32 console project
Set name and path
Select an empty project in the next panel
Drag and drop source file into solution explorer

Next to run openMP

Goto Project properties -->configuration properties/C.C++/language
Activate the openMP tab

Check out the link

Linux

To install and try out openMP (if you have the gcc compiler already) you have to do ....nothing
If you are using gcc as the compiler then you have done it already. All there is to do is create a c/c++ code and while compiling the program simply use the following statement

gcc -fopenmp c_code.cpp -o c_obj

where c_code.cpp is the source code and c_obj is the object /exe file to run

Parallel Programming (OpenMP, CUDA, OpenCL)

One of the various useful tools which I feel will be required if we need to do some simulations, applications, etc.. to run efficiently. But the latest computer processors from intel (corei3, corei5, corei7) have multiple cores so isn't it implied that we are getting better performance ?

The short answer is NO

They simply have the functionality however if we are writing a code which performs sequential actions then we are not utilizing the resources efficiently.

We can see that if some computations are independent then they can be performed in parallel using threads and this reduces the computation time of the overall program drastically.

To learn openMP a YouTube_series is provided by intel and instructor is Tim Mattson. The Basics of openMP and a few parallel algorithms are taught in C/C++ and fortran.

A series called Heterogeneuous Parallel Programming is also offered in Coursera which teaches basics of CUDA and openCL .

Raspberry Pi

The Raspberry Pi could be called a "Mini Computer" as it is very cheap and useful. On purchase of a Raspberry Pi you would have to purchase/get a mouse+keyboard, any adapter to connect you to the internet, a HDMI cable+ Display with HDMI connection. It works well with the Linux OS Raspberian (Raspberry Pi + Debian) but other distros can be used such as fedorapi.

Tips to efficiently do computations

Hint #1

You can configure eclipse in Linux to use as a debugger.
To configure with eclipse for c/c++ :

Download the HEVC reference software version using subversion (First check if files such as .cproject, .project, etc exist) and later zip the file
Open Eclipse and select file and select import. Select the zipped file and give a name to your Project.
The Zip file will be unzipped at the workspace location.Goto the build linux location i.e. ‘Project Space’/build/linux/ and enter the terminal command “make”. This will create the binaries and object files.
After that go to the project properties,select c/c++ build,select builder settings ensure that the “Use default build command”is unchecked and type “<Project Space>/build/linux/makefile”. Ensure the Build Location is also in the same directory
Since there are many project you have to first select the project of interest e.g. TAppEncoderStatic Check the projects name as well
In the Arguements section type the necessary arguments e.g. configuration files -c [config file] (Note ensure that they are in the bin directory along with the YUV file of choice ). Also ensure the workspace is also the same directory

Hint #2

Depending on various factors such as number of frames and resolution of the input video the processing time may vary (usually an hour). To effectively utilize the system and time, simply open multiple command prompts(on windows)/terminals (on linux). This is called High Throughput Computing (HTC) where the same program is executed but with different parameters.
Bash programming can also be used to automate processes. Literally making a bash script is easy by adding the bang line #!/bin/bash
Python is also an alternative to bash scripts as it can have problems while defining arrays. Utilize HTC wherever possible but do not push it as PCs have certain capacities and capabilities.

JM and JSVM software are the reference softwares used to understand and emulate H.264 and its scalable version while HEVC and SHM software are the reference softwares used to understand and emulate H.265 and its scalable version. These reference softwares( JM, JSVM, HEVC, SHM ) are managed using a version control software such as subversion (SVN) and Concurrent Version Systems (CVS). These applications help in controlling and distributions of the reference softwares.The HEVC and its scalable version reference software can be downloaded
from

https://hevc.hhi.fraunhofer.de/

Windows

In windows one can use a simple software such as tortoiseSVN or tortoiseCVS and simply get the corresponding tools. Also some standard YUV files will be required to test the code such as foreman.yuv,basketballdrill.yuv etc.. To read and play the yuv files in windows ,YUV player or YUV toolkit (preffered) will be useful. HEVC uses SVN so for windows(after installing tortoise SVN) go to the folder where the project will be saved.Simply right click and select the option “SVN checkout”and add the url and click OK.
https://hevc.hhi.fraunhofer.de/svn/svn_HEVCSoftware/

Linux

In Linux,the packages “subversion”and “cvs ”have to be installed via command line or any other package manager(e.g. synaptic package manager). After installation ,in the terminal , simply type the command in the terminal window:

svn checkout https://hevc.hhi.fraunhofer.de/svn/svn_HEVCSoftware/

Unfortunately YUVplayer/YUV toolkit is not developed for linux however we can use the VLC player to read these YUV files. To read and play the yuv files in linux using vlc player use the command.

vlc --demux rawvideo --rawvid-fps FrameRate --rawvid-width FrameWidth --rawvid-height FrameHeight --rawvid-chroma I420 YUV_File

After the download is complete in either platform, a set of directories/folders will appear in your workspace such as:

build (which has the project files )
cfg (configuration files)
compact
doc (documentation)
HM.xcodeproj
lib (libraries)
source (source files)

The“build”contains the projects files like .sln for visual studio projects (in Windows)and Makefiles (in Linux). The“source” folder contains all the C/C++ source files,headers and libraries required the execute the code. The“cfg”folder contains the configurations files which are generally used to avoid inputing large command line arguments at the time of execution. Once
the files have been compiled another folder “bin”will appear which will contain the binaries i.e. the executable files. Depending on the system configurations (i.e processor, RAM ,HDD size etc..) the time required to complete the execution may vary. It is recommended to use a system with good configurations.

Windows

In windows,the visual studio express(preferred 64 bit)software is used for compilation of the project.The Project files are present in the build folder with a .sln extension. While opening the visual studio project, a set of projects will be displayed on the solution explorer.First the solution of these projects must be built.
To build the solution of these projects simply click F7 or right click and select build solution and the executional files will be created. Now to select a particular project e.g. the encoder project “TAppEncoderStatic” first right click on the project and select ‘Set as Startup Project’. Next right click on the project and select properties go to configuration properties and select the
debugging tab.The Command arguments are already added however if not simply check the software manual in section 3.

Using the encoder it will show that has to be written. Command Prompt can also be used by going to the directory and using the command

TAppEncoder.exe [-c config file] [--parameter value]

E.g. TAppEncoder.exe -c BasketBallDrill.cfg -c encoder_lowdelay_main.cfg

Using this command the following outputs should appear if not then an error has occurred.

Linux

In Linux, open the terminal and go to the linux directory and type the command ‘make’. It will compile the code and generate executable files.These executable files can be used in the terminal by using the following command

./TAppEncoder [-c config file] [--parameter value]

E.g. ./TAppEncoder -c BasketBallDrill.cfg -c encoder_lowdelay_main.cfg

Scalable Video Coding

High Efficiency Video Coding (HEVC) is relatively new video coding standard (H.265) which has been released in January 2013. This standard augments the already existing standard (H.264) to provide enhanced quality such as high definition (HD) and ultra high definition (UHD) services with a reduced bit rate. This increases the amount of processing required to encode/decode the bit stream. For implementation, understanding of the codec in the industries, academic domain a reference software has been released to understand the functioning and experimentation.

Android Setup

Its very easy to start with android programing as the programming language is a combination of Java (with some additional libraries for back end processing) and XML (for the front end Graphical User Interface GUI). All that is required is that java is installed on the operating system used(Windows, Mac, Linux, etc...). After Java has been successfully installed and configured(setting the path variable on the OS) the next step is to simply download either of the following:

Android Development Tools (ADT) bundle (the complete package ,highly recommended as everything is already set however no longer supported)
Eclipse IDE for java developers(after helio version )+android Software Development Kit (SDK)+android adt plugin (all downloaded separately and have to be configured which is a time consuming process)
Android Studio which is an IDE provided by google based on IntelleJ (a complete package)

After acquiring the software, the next task is to download the tools required (such as the image file ) for the android version (Lollipop ,KitKat, Jellybean etc..) from the android sdk manager.

Android Programming + openCV Libraries

As the name suggests,we can combine openCV code written in C/C++ and Android code which is written in Java.You may ask yourself how is it possible as Java and C++ use different compilers for compilation and how can they be called by each other.

The simple answer which explains it is the android Native Development Toolkit (NDK).

There are new java libraries so it isnt necessary to embed C++ as well but it may be useful at time for code reuse-ability.

Sources:

Android Development

Android Developement has been the latest buzz ever since the sdk was first released for developing android apps for free and understanding the fact that we can develop any app we require and correspondingly upload it into the Android Market (Google Play).

I will be sharing my experience of android developement and revealing my sources.

Major thanks to #mybringback and #thenewboston by travis and bucky as they guided me step by step to develop simple apps and test them easily.

Sources: