You are not logged in.
Hey everyone,
Check out this contest, there's a 2000$ cash prize: http://www.dermandar.com/puzzle.zip
Deadline 30 march 19h00 Beirut time.
Good luck :)
I imagine at least 3 people from this forum will be able to solve this problem. Then what? Timing based?
From the puzzle's readme:
The best program (elegant, simple, fast) will be awarded 2000$.
I suppose these parameters are taken into consideration: the speed of the algorithm's execution, the algo itself (brute force vs "smart" algo), the code's readability and cleaningness (regardless of the language)... etc
This is not so much a contest, as it is recruitment.
Getting good talent these days is virtually impossible and the best way to find those, are through such 'contests'.
Yes the prize is real, 2000$ to identify good talent is nothing.
@xterm
we see it as looking for talent, others see it as 2000$ in 3 days...
what happens after it is totally random. winners aren't forced to anything!
@xterm
we see it as looking for talent, others see it as 2000$ in 3 days...what happens after it is totally random. winners aren't forced to anything!
Oh I'm sorry, don't get me wrong. This is perfectly acceptable!
Good luck!
I think its a good idea to find talent, good luck ^_^
That's easy, I did something similar before... Too bad i'm busy in the next few days
I suck i can't do this.
Border images have a duplicated row/column of pixels.
I am assuming this means a top border for example would have the same set of pixels on the top and top minus one lines of pixels. ie:
new Rectangle(0,0,52,1) and Rectangle(0,1,52,1) should have the same set of pixels for this to be a top border.
(where Rectangle(x,y,width,height) in the bitmap coordinates.
This is what I could understand from the problem [text], which is only giving me 46 top border images instead of 80.
So I guess I am not understanding the problem :)
I think we should get the first column or row of pixels and compare it to the other raw images column or row and check for the number of matching pixels, RGB of course, of the following raw images. The more number of matching pixels available, the higher the probability that this raw image is the upcoming one of the puzzle.
I'm not sure thought, this is what i think should happen.
Each adjacent pair of images overlap on 1 row/column of pixels. Border images have a duplicated row/column of pixels.
Just spent one hour re-doing image manipulation to check if there was a problem in my logic, but it is fine, I am only able to find 46 top borders and 49 bottom borders (instead of 80 x 25). which is really weird.
Can you please explain the above quoted sentence?
Border images have a duplicated row/column of pixels.
This definitely needs a better explanation.
Are you referring to the 4 corner images only ? or the entire row/column ?
her's a better explanation hopefully
http://i.imgur.com/SzVu7.jpg
off topic: Hey you have a galaxy note?
There is no reference to the contest on the website?
@rolf no
@elzalem can you please double check the data?
none of the tiles have the same leftmost column pixel values equal to its topmost row pixels as per the illustration
@ZeRaW, display the tiles to see things clearly.
I just discovered this contest... and now I coded some lines in C and I only got 49 as top blocks while I guess It must be 80...
I'm able to display the tiles clearly with no problems... I'll try to see what's wrong and report back to you...
I just discovered this contest... and now I coded some lines in C and I only got 49 as top blocks while I guess It must be 80...
I'm able to display the tiles clearly with no problems... I'll try to see what's wrong and report back to you...
are you sure you are not flipping rows and columns?
@geek, I did this in the first step, (80 x 25 tiles) which results in a 4000 x 1250 image.
Sadly from the sktech, the top left tile where the image starts should have its left column of pixels equal to its top row of pixels and non of the tiles passes this test.
So even my greedy algorithm fails which is so recursively simple.
454447415244 wrote:I just discovered this contest... and now I coded some lines in C and I only got 49 as top blocks while I guess It must be 80...
I'm able to display the tiles clearly with no problems... I'll try to see what's wrong and report back to you...are you sure you are not flipping rows and columns?
@geek, I did this in the first step, (80 x 25 tiles) which results in a 4000 x 1250 image.
Sadly from the sktech, the top left tile where the image starts should have its left column of pixels equal to its top row of pixels and non of the tiles passes this test.
So even my greedy algorithm fails which is so recursively simple.
It doesn't matter if you are flipping the row or not, at the end, you might get a flipped BMP image, but that won't be a problem.
In a BMP image header, if ImageHeight was negative then the image will be treated from top to bottom. Anyway... I think you assumption is wrong. The top left image should have the first and second rows equal + the first and the second columns equal too...
@454447415244:
you are right, I was assuming that the top left would have the top and top-1 rows equal and left and left-1 columns equal but the sketch done does not illustrate that so i thought he meant the left would be equal to the top.
additionally these are not BMPs so no headers. my main concern was that I was counting all the tiles/images that are top borders and all the tiles that are bottom borders which should give me an equal number of 80.
but it is actually giving me 46 top bordered tiles and 49 bottom bordered tiles. Please note I am using direct value comparison without any kind of threshold.
Last edited by ZeRaW (March 28 2012)
@ZeRaW, it seems you need to do more research in image processing field.
I don't want to spoil the challenge. Good luck!
The settings to open the raw blocks in Photoshop are:
52x52 pixels, 8 bit, 3 channels, interleaved, header size = 0
The files are padded at the end with 51bits of empty bits (all 0)
Last edited by rolf (March 28 2012)