I would probably try open CV as that has a lot of tools for computer vision tasks
#Making an algorithm count the pieces from a pdf manual
south linden
flint pawn
There are libraries for reading the PDFs, but you'll have to put in some extra work for actually counting the pieces. Could you send some examples of what they look like? It's been a while since I've last touched lego but I guess you could crop each of the pages to only the rectangle in the corner containing the bricks needed for that step, then you would have to come up with an algorithm to split this list of pieces up to get the individual bricks as well as their count.
Once you have that it should be fairly easy to group them (e.g. by image similarity) and sum their counts
sly apex
This is one of the pages
flint pawn
So we can't just cut out rectangles for the parts but at least the numbers don't overlap
How many pages do you have? I have a feeling that Python might not be fast enough here if you have lots of manuals
sly apex
The manual has 206 pages
I am sorry to ask, but what exactly does "building from source" mean?
flint pawn
compiling the source code of that library
sly apex
What other language could I use instead of python? Does java perhaps have a library similar to open cv? I found one for javascript, but I am wondering about its speed to solve that
flint pawn
What other language could I use instead of python? Does java perhaps have a libr...
I think Python will be fine for 200 pages, can you send me the manual? Kind of tempted to rewrite it in Rust (™️)
viral reefBOT
@sly apex
flint pawn
🗿 You can just send me a direct link or something as well
trim wraith
This is one of the pages
from what i understand you want the info in the little blue box, are looking for just the total amount of pieces or how much of each piece is required?
sly apex
How much of each piece is required
trim wraith
How much of each piece is required
and how are you planning to index it? like picture:amount required, or piece_name:amount required?
sly apex
With the pictures
flint pawn
Step one: Figure out how to read the PDF in Python and then count the pages (so we can see you read it properly)
sly apex
Sir yes sir! O7
flint pawn
😈
trim wraith
With the pictures
then you can do this with the code, you can subsection the image to only look at the blue box by scanning for a square that contains a certain color.
once you get the square you can subsection each piece, you can do that by line scanning, you create an empty array of images, with them will also be the polygon shape they are. you scan through each line, if the point is not the blue color you add it to the new image your making, you do that with each pixel, once the line scanner reachs blue again (meaning that it reached the end of the object), you go to the next line and repeat until the entire line is blue then add that image to the array as well as the polygon shape of it. to make sure your not getting a block you already scanned, you check the point in the list of polygon shapes you made to make sure its not in there. then do that with each block until its not adding anymore blocks to the array. then you need to scan the text, get each block position then each text position and just find the text that is closest in the vertical axis
sly apex
Would it also work, if I use the color of the rectangle's outline?
