Dasara Dairies-2015: Yelahanka Sri Vasavi Devi Temple

Dasara days are really fun filled days, where we get to see all people roaming around who are eager to see all different alankarams being done in different  temples, Really appreciate the patience of archaks to nicely decorate the idols of GODs with their creative ideas. Also we should thank the temple authorities as they would plan and provide all necessary things to them.

This post wouldnt have been more colorful without the help of Yelahanka Vasavi Yuva Vedike team, a big thanks for providing the photos of Sri Vasavi Devi Temple.

Dasara Invitation

Initially as I mentioned, the temple authorities play a very crutial role in pre-planning all the work, Just to give a simple example, we can take a look at the below image, which they have released, much before dasara started about all the different alankaram schedule.etc.

List of Alankarams

Day 1 (13-10-2015): Ganesha-Sugarcane; Vasavi Devi-Sri Subramanya Swamy; Sri UmaMaheshwari-Avalakki; Sri Lakshmi Narayana-Kundan

Day 2 (14-10-2015): Ganesha-Caudury Gems; Vasavi Devi-Bala Tripura Sundari; Sri UmaMaheshwari-Biscuit; Sri Lakshmi Narayana-Chocolate

Day 3 (15-10-2015): Ganesha-Muttu; Vasavi Devi-Sri Narashima Swamy; Sri UmaMaheshwari-KadleKaalu; Sri Lakshmi Narayana-Carrot

Day 4 (16-10-2015): Ganesha-Pancha Varna; Vasavi Devi-Sri Kapileshwara; Sri UmaMaheshwari-Grapes; Sri Lakshmi Narayana-Flowers

Day 5 (17-10-2015): Ganesha-Beetle leaves; Vasavi Devi-Lord Venkateshwara; Sri UmaMaheshwari-Vhibhuthi; Sri Lakshmi Narayana-Sri Ranganatha Swamy

Day 6 (18-10-2015): Ganesha-ChittaVastra; Vasavi Devi-Sri Saraswathi Devi; Sri UmaMaheshwari-Sri Gandha; Sri Lakshmi Narayana-Grapes

Day 7 (19-10-2015): Ganesha-Benne; Vasavi Devi-Matsyavatara; Sri UmaMaheshwari-PanchaVarna; Sri Lakshmi Narayana-GejjeVastra

Day 8 (20-10-2015): Ganesha-GejjeVastra; Vasavi Devi-Sri Vamana; Sri UmaMaheshwari-Kundan ; Sri Lakshmi Narayana-Vhibhuthi

Day 9 (21-10-2015): Ganesha-Grapes; Vasavi Devi-Sri KhaLikambha; Sri UmaMaheshwari-Coconut; Sri Lakshmi Narayana-Muttu

Day 10 (22-10-2015): Ganesha-Vegetables; Vasavi Devi-Sri Santana Lakshmi; Sri UmaMaheshwari-DryFruits ; Sri Lakshmi Narayana-Sri Gandha

Day 11 (23-10-2015): Ganesha-BeLLi Kavacha; Vasavi Devi-BeLLi Kavacha; Sri UmaMaheshwari-Rudrakshi Flowers; Sri Lakshmi Narayana-Harishina Kunkuma

And at the last but not least we should also thank all the sevakartas as well, which is very important and I wish every year more number of people come forward to become sevakartas.

Is strtok safe !?

We all know that strtok is the most convenient way we all use to tokenize strings.

Syntax:
        #include <string.h>
        char *strtok(char *str, const char *delim);
        where,
        string to be tokenized
        delim - set of delimiters to be used while tokenizing

Example:

#include <iostream>
#include <string.h>
using namespace std;

int main()
{
    char paragraph[] = "Hello$Funny$Man"; // String where words a seperated  by $ 
    // We need to tokenize above string
    char* tok;
    tok = strtok(paragraph,"$"); //  extract the word using $ as a delimiter
    while(tok)
    {
        cout << "Extracted word is " << tok << endl;
        tok = strtok(NULL,"$");
    }
    return 0;
}

Output: 

Extracted word is Hello
Extracted word is Funny
Extracted word is Man

Now is it the safest way to tokenize string/char-array? Do we see any problem with this method ?
What happens when we have to sub-tokenize the extracted token ? Let see with the below example.

#include <iostream>
#include <string.h>
#include <stdlib.h>
using namespace std;

int main()
{
    char paragraph[] = "Hello$Funny$Man How$are$you$doing?"; // String where sentence a seperated by space and words by $ 
    // We need to tokenize above string
    char* tok;
    tok = strtok(paragraph," "); // Firstly extract the sentence using space as a delimiter
    int count = 0;
    while(tok)
    {
        char* subtok;
        char* sentence = strdup(tok);
        subtok = strtok(sentence,"$");
        count++;
        while(subtok)
        {
            cout << "Extracted word in sentence " << count << " is " << subtok << endl;
            subtok = strtok(NULL,"$");
        }
        free(sentence);
        tok = strtok(NULL," ");
    }
    return 0;
}

Output: 

Extracted word in sentence 1 is Hello
Extracted word in sentence 1 is Funny
Extracted word in sentence 1 is Man

We can observe that only one sentence was extracted, this happens because, if we closely observe the usage of strtok, strtok takes source string to be tokenized only in its first call, from the second call on-wards it takes NULL as the first argument, which means strtok uses some global scope space to store the pointer(point to which it has tokenized)

If we use strtok both in outer loops and inner loops, it only runs the outer-loop once as we saw in above example. Since there is only one common global pointer and by the time inner loop is completely executed, this global pointer points to NULL, and hence the actual outer loop pointer(point to which it has tokenized) is lost.Therefore it exits out of outer loop assuming it has tokenized completely.

How do we solve this problem?? well we don't we to do anything, this is a open secret which is in the man page of strtok.
string.h also has an another API called strtok_r which takes in an additional parameter to store the context of the string/char-array i.e point to which it has tokenized.

Syntax:
        #include <string.h>
        char *strtok_r(char *str, const char *delim, char **saveptr);
       where,
        str - string to be tokenized
        delim - set of delimiters to be used while tokenizing
        saveptr - additional pointer to save the context

As per man page its strtok_r() function is a reentrant version strtok(). The saveptr argument is a pointer to a char * variable that is used internally by strtok_r() in order to maintain context between successive calls that parse the same string.

Lets us use strtok_r and see if it solves above problem or not.

#include <iostream>
#include <string.h>
#include <stdlib.h>
using namespace std;

int main()
{
    char paragraph[] = "Hello$Funny$Man How$are$you$doing?"; // String where sentence a seperated by space and words by $ 
    // We need to tokenize above string
    char* tok;
    char *savePtr1, *savePtr2;
    tok = strtok_r(paragraph," ", &savePtr1); // Firstly extract the sentence using space as a delimiter
    int count = 0;
    while(tok)
    {
        char* subtok;
        char* sentence = strdup(tok);
        subtok = strtok_r(sentence,"$",&savePtr2);
        count++;
        while(subtok)
        {
            cout << "Extracted word in sentence " << count << " is " << subtok << endl;
            subtok = strtok_r(NULL,"$",&savePtr2);
        }
        free(sentence);
        tok = strtok_r(NULL," ", &savePtr1);
    }
    return 0;
}

Output:

Extracted word in sentence 1 is Hello
Extracted word in sentence 1 is Funny
Extracted word in sentence 1 is Man
Extracted word in sentence 2 is How
Extracted word in sentence 2 is are
Extracted word in sentence 2 is you
Extracted word in sentence 2 is doing?

From the above output we can observe that the problem is solved as both the sentences are tokenized. So its always best practice to use strtok_r when you know that your program is going to use strtok in recursions or threads or nested-loops or to put the other way around would be use strtok only if you are damn sure that your program doesn't use strtok in resursions or threads or nested-loops.

Happy Coding !!

Scientific Notation - best way to print floating point numbers

Dealing with floating point numbers is one of the most complex thing which a programmer and computer deal with in a day-to-day programming.

Floating point numbers are real numbers which has a fractional part.
For Eg: 1.234, -2.345 .. etc

As we all know that computers are the integer based machines, [Basically it can understand only 2 voltages high (1) and low (0)], and there are complex ways which includes lots of assumptions and approximations in representing floating point numbers, In this post i will only focus on the best way to print floating point numbers.

There are few standards set by IEEE for using floating point numbers. So coming to printing of floating point numbers on the standard o/p (monitor) or in the file, which is the best practice to be followed ? Which types of notation is more precise and more consistent.... Answer to that would be scientific notation

In scientific notation (which is considered most standard way) any number can be represented in the form :

x*10^y

For Eg : 0.2 --> 2*10^-1
              200 --> 2*10²


The following c++ code and its output could prove to be a simple proof to explain why is scientific method the best way to print floating point numbers.

// Non-scientific way
#include<iostream>  // To print to standard output
#include <iomanip> 
using namespace std;

int main () {

    double shift = 10616.200000000001;
    double num1 = 0;
    double num2 = 0.01;
    
    std::setprecision(15);
    // cout << scientific;

    cout << "(shift + num1) * 1e-3 = " << setw (12) << (shift + num1) * 1e-3 << endl;
    cout << "(shift + num2) * 1e-3 = " << setw (12) << (shift + num2) * 1e-3 << endl;

    return 0;
}

// Output
(shift + num1) * 1e-3 =  10.6162
(shift + num2) * 1e-3 =  10.6162

From above output we can observe that even tough we had set the precision to more than what we need, final result is same irrespective of num1(0) and num2(0.01) begin different.

Now let us try to print the same numbers in scientific way and see the difference. We can just use cout << scientific to print in scientific way. From the above code we need to just remove the comment line.

// Scientific way
#include<iostream>  // To print to standard output
#include <iomanip> 
using namespace std;

int main () {

    double shift = 10616.200000000001;
    double num1 = 0;
    double num2 = 0.01;
    
    std::setprecision(15);
    cout << scientific;

    cout << "(shift + num1) * 1e-3 = " << setw (12) << (shift + num1) * 1e-3 << endl;
    cout << "(shift + num2) * 1e-3 = " << setw (12) << (shift + num2) * 1e-3 << endl;

    return 0;
}

And its output looks like this.

//Output 
(shift + num1) * 1e-3 = 1.061620e+01
(shift + num2) * 1e-3 = 1.061621e+01

We can very clearly observe that the result is different while printing in scientific notation.


Hence scientific notation is the best way as it does not lose more information due to approximations and its safe way to deal with floating point numbers.

A very big thank you to http://hilite.me/ which helped me beautify the code.

Magic Squares

Was just remembering my childhood days this diwali, during childhood which we used to have lots of fun with friends and used to play a lot.

Suddenly remembered that trick which we used to use to fool around with friends, and that's magic-squares.

Magic-Squares is just arrangement of numbers in the square shape where, sum of each of its rows and sum of each of its column and sum of each diagonal is equal to same number.

So now being a computer science enthusiastic, I thought ok.. let me tell my computer (which had g++ compiler) to solve magic-squares.

Here is the result of my computer did. It did all these below ones in less than a fraction of a second ;) :P

NOTE:
After the symbol => is the rowSum
After the symbol \/ is the columnSum
RTLB -- Sum of numbers across Right Top to Left Bottom Diagonal
LTRB -- Sum of numbers across Left Top to Right Bottom Diagonal

Well for those who are interested in having a look at the code, here it is for them.

This is our main function:

 int main()  
 {  
   int num = 0;  
   cout << "Enter any odd number: ";  
   cin >> num;  
   cout << endl << endl;  
   if(num % 2 == 0) {  
     cout << "Even number not supported !! " << endl;  
     return 1;  
   }  
   MagicSquare mSq(num);  
   mSq.computeMagicSquare();  
   return 0;  
 }  

This is the main part where the magic square is begin populated.

   void getRowAndColumn(int& row, int& col, const int counter)  
   {  
     if(counter == 1) { col = num / 2; return;} // Initial condition to insert 0  
     else if(row == 0 && (col == num / 2)) { row = num - 1; col++; return;}  
     else if(col+1 == num && row == 0) { row++; return;}  
     else if(col+1 == num) { col = 0; row--; return;}  
     else if(row == 0) { row = num -1; col++; return;}  
     else if((row!=0) && (col!=num-1) && magicSquare[row-1][col+1] != 0) { row++; return;}  
     else { row--; col++; return;}  
   }  
   void computeMagicSquare(void)  
   {  
   // Printing Dummy Magic Square  
   // Solution for magic square   
   // Step 1 : Placing 1 in the middle of the first row 1st row of the matrix  
     cout << "Printing matrix template:" << endl;  
     printMagicSquare();  
     int counter = 1;  
     int row = 0;  
     int col = 0;  
     while(counter <= num*num)  
     {  
       getRowAndColumn(row, col, counter);  
       // cout << "Row = " << row << "Col = " << col << "Counter = " << counter << endl;  
       if(row < 0 || row >= num || col < 0 || col >= num) cout << "Error in calculating Row/Column Row = " << row  
         << "Column = " << col << endl;  
       magicSquare[row][col] = counter++;  
     }  
     cout << "\n\nPrinting magicSquare: " << num << "*" << num << endl;  
     printMagicSquare();  
   }  

This is the printing portion of the magicSquares.

   void printMagicSquare(void)  
   {  
     int LTRBDiagSum = 0;  
     int LBRTDiagSum = 0;  
     int *columnSum = new int[num];  
     for(int row = 0; row < num; row++)  
     {  
       int rowSum = 0;  
       for(int col = 0; col < num; col++)  
       {  
         rowSum += magicSquare[row][col];  
         if(row == col) LTRBDiagSum += magicSquare[row][col];  
         if(row+col == num-1) LBRTDiagSum += magicSquare[row][col];  
         columnSum[col] += magicSquare[row][col];  
         cout << "\t" << magicSquare[row][col];  
       }  
       cout << " => " << rowSum << endl;  
     }  
     cout << "RTLB(" << LBRTDiagSum << ")";  
     for(int i = 0; i < num; i++)  
     {  
       cout << "\\/";  
       if(i != num-1) cout << "\t";  
     }  
     cout << "LTRB(" << LTRBDiagSum << ")";  
     cout << endl;  
     for(int i = 0; i < num; i++)  
     {  
       cout << "\t" << columnSum[i] ;  
     }  
     cout << endl;  
     delete[] columnSum;  
   }  

And finally our class looked like this

 class MagicSquare {  
   public:  
   MagicSquare(int n)  
   {  
     num = n;  
     magicSquare = new int* [num];  
     for(int i = 0; i < num; i++)  
     {  
       magicSquare[i] = new int[num];  
     }  
   }  
   ~MagicSquare()  
   {  
     for(int i = 0; i < num; i++)  
     {  
       delete[] magicSquare[i];  
     }  
     delete[] magicSquare;  
   }  
   void printMagicSquare(void);  
   void getRowAndColumn(int& row, int& col, const int counter);  
   void computeMagicSquare(void);  
   private:  
   int **magicSquare;  
   int num;  
 };  

Well for those is really interested to execute this source code at your end, you can get the code from here.

I sincerely thank this blog for helping me formatting above code snippets.

Yeah thats it about magic-squares. Lets see if time permits will try to enhance code for even numbers as well.

Weekend trip from Bengaluru

Bengaluru--Sringeri--Agumbe--Udupi--Dharmastala--Kukke--Bengaluru

Traveling is amazing if you have a good company. This time it was with 2 of my cousins. We just thought bangalore is heating up very much this summer and want to beat the heat by hanging out in some nature's beauty where we can chill out.

So meet the crew :

Nagarjun(L) with Kiran(and thats myself;) )(R)

Nagarjun(L) with Shreyas(R)

Plan :

Initial plan was to leave bangalore on friday evening and then visit Sringeri early in the morning, witness the temples and holy places there, later go to Agumbe and enjoy nature's beauty, which happpend accordingly and the rest of the tour was dynamically planned.

We hired a auto-guy in agumbe who helped us roam around that place.(That place was real nature's beauty)

Meet this auto-guy. (Took his permission to take his photo)

And his auto with his phone number (Took his permission to take this photo as-well)

Here are some of the photos which we captured, Nagarjun did take a lot of photos which i am missing now while writing this post.

The view from the top of the hill was breath-taking.

This is the very ancient temple on the hill.

Looks small but very very deep well.

Amazing is'nt it!! Looks like one of those children slides

Nagarjun and Shreyas busy boating !! (This photo was taken while boating so the photo quality got messed-up;))

After witnessing the world famous sunset point in Agumbe, we headed towards Udupi which is 55 Kms form Agumbe, there are of plenty local mini buses which ply from nearby places to Udupi via Agumbe.

Udupi . We visited Sri Krishna Temple and Madhwa Charya Matha there.
We got chance to see elephant moving the chariot of Lord Sri Krishna and then after evening Aarthi at 8.30 pm we went inside the temple for darshnam, we had a very good darshan and it was very blissful experience to see Lord Sri Krishna of Udupi, At that time of darshan, i was remembering the Madhurastakam which has the line in that as "...Udupiya Sri Krishna ..." It was a very good experience. On the day we visited it was Ekadashi, so we missed the chance to have dinner in temple.

Then next day we woke up early in the morning by 4.00 am and again went to temple to see Subrabatha seva to Sri Lord Krishna and we have again had a very good darshnam and also got a chance to see Abhishekam being perform to Lord Krishna.

Later we headed towards Dharmastala. Saw very famous Sri Dharmasthala Manjunateshwara Temple there. We had to stand in queue for quite some time there as we went during Maha neiveidya timings 11 - 12 noon.

There are many buses with lots of frequency from Dharmastala to Kukke. Went to Sir Kukke Subramanya Swamy temple and it was very pleasant evening, really had a very peaceful time.

We had a return booking from Kukke to Bangalore in KSRTC Airawat we just got in to the bus and by the time we opened our eyes we were in Majestic, Bangalore
Back home :)

Dasara - Navratri Alankaram's (Temples in Yelahanka)

Dasara Times - Temple decorations in Yelahanka-Oldtown

Ardha-Nareshwar Alankaram (Incarnation of Shiva-Parvathi) [Vasavi Devi Temple]

PanchRangi Alankaram (Alankaram with colors) [Yelamma Devi Temple]

Saraswathi Alankaram [Vasavi Devi Temple]

Shiva-Parvathi in Pancharangi Alankaram (Alankaram with Colors) [Vasavi Devi Temple]

Shiv Ji (Har Har Mahadev) [Vasavi Devi Temple]

Lakshmi Narayana in Sindhur-Carpoor Alankaram [Vasavi Devi Temple]

Ganesh ji in Pancharangi Alankaram [Vasavi Devi Temple]

Swaamiye Sharam Aiyappa [Muneeshwara Swamy Temple]

Saraswathi Alankaram [Chowdeshwari Devi Temple]

Subramanya Swamy in Chandan Alankaram [Chowdeshwari Devi Temple]

Ganesh ji in Benne (Butter) Alankaram [Chowdeshwari Devi Temple]

Sri MahaVishnu [ Muneeshwara Swamy Temple]

Durga Devi Alankaram [Chowdeshwari Devi Temple]

Sri Parvathi Parameshwara in Chandan Alankaram [Vasavi Devi Temple]

Durga Devi Alankaram [Vasavi Devi Temple]

Hatti (Cotton) Alankaram [Vasavi Devi Temple]

Drakshi Godambhi (Resins and Cashew) Alankaram [Chowdeshwari Devi Temple]

Gayatri Devi Alankaram [Vasavi Devi Temple]

Lord Venkateshwara Alankaram [Chowdeshwari Devi Temple]

Vajranghi Alankaram [Sri Ganesh Temple]

Droupadi Devi Alankaram [Maheshwari Devi Temple]

My journey with face recognition

Hello friends,

Today i actually wasted around 2 hours of time in the morning for a small bug in some of the images of yale data set. And I don't want any other person to waste time on the same thing so i taught of writing this post.

Before actually going into the details of that, let me tell about my journey with face recognition.

As early as in my 4th semester, I approached one of the professor in my college to help me do any project. He was actually pursuing his phd in the field of image processing. His interests were on face recognition. Every meeting i had with him, he actually tried to create some interest on that field in me. That was finally indeed very helpful to me in doing my final sem project. Thanks to our Prof.GTS

Face recognition is actually very interesting field of study and its one of the most interesting topic in the field of computer science research and development.If you want to know more on this topic, please visit the renowned official website of face recognition community.

Usually many people in this field use java or matlab for coding purposes and some others use shell scripting such as Image Magiks and some others use c or c++ or cv (computer vision modules).

Like every other person even i started off with matlab as said by experts in the domain, I actually have no issues with matlab it's cool and user friendly, its actually the most easiest tool which any one can find on this planet for doing coding of face recognition algorithms.

But one of my good friend told me use python language for coding. During the initial stages i actually totally ignored him saying matlab is very good and i have done all the hard work learning matlab. But he was not convinced with me, he actually said that learning python just for sake of atleast doing project for final semester is worth then being a expert in matlab, we even had heated discussions many times regarding matlab or python.

But at last i taught alright, let me see whats there in python and i started using python for face recognition, its one of the easiest programming language any one can learn on the planet, It's actually one of the best programming language i have ever seen. After seeing python i actually felt why did i wasted all my time learning matlab and i literally felt that i should have listened to my friend long time back, it was not very late "Better late then never". Thanks a lot to my dear friend KG.

In brief, in face recognition what happens is some set of images is trained and mapped into what ever the way algorithms wants to. When it comes to training a test image image is mapped onto the mapped trained image space and euclidian distance between the test images and all the trained images are taken, the image corresponding to the minimum euclidian distance is the recognized image.

I would just like to tell one advantage using python over matlab, there are many advantages tough.

(1) Python is open source and it actually takes around 10 MB of your disk space. It doesn't demand for high end machines, on the other side matlab itself is some 2 GB of size and it actually demands for very high end machine.
(2) Using python writing generic code, writing one small piece of code for all the databases is very easy and in matlab we actually have to write separate matlab-files for individual databases.

Python has a very good support for image processing. Please check out python packages - scipy,numpy,PIL,Image for details of the support of python in image processing.

Let me come back to why actually did i waste the time today even tough it wasnt my mistake.

Yesterday i downloaded yale dataset ( one of the standard dataset for experimental purposes ) in the .pgm format from here .

When i started to test this dataset i was actually getting some strange error, look at the following screen shot, some knowledge in python will actually makes you understand what that it say.



It actually says, i cannot load some image blah blah etc.., Thats was indeed a very strange thing i encountered because i had loaded other images from the same database successfully. After doing some experimentation i was able to find that some images in yale database which i downloaded from here were the faulty images. Yes i realized that this was the reason why two similar databases were stored in that same server. You can actually find the proper database here with the folder named yalefaces(copy) , I actually dint wanted any other person to waste time on the same issue so i taught of sharing this experience.

If in case you are stuck up with the same problem, i think this should be of some help to you. I actually got the above error when i tried to load the following images of yale database.

1. subject11.normal.pgm
2. subject09.rightlight.pgm
3. subject06.sad.pgm
4. subject13.rightlight.pgm
5. subject10.noglasses.pgm
6. subject15.rightlight.pgm
7. subject06.happy.pgm
8. subject13.leftlight.pgm
9. subject08.rightlight.pgm
10.subject03.rightlight.pgm
11.subject10.sad.pgm
12.subject04.rightlight.pgm
13.subject10.wink.pgm
14.subject14.leftlight.pgm

But the same images works well when i downloaded from the yalefaces(copy)

In case if you want any help regarding implementation of face recognition algorithms in python please feel free to contact me, i will give my best shot in helping you out.

Finally one thing i wanted to say is Python rocks \m/

Thanks a lot dear reader :)