Image Recognition and Python Part 7

Key Takeaways

hello everybody Welcome to part seven of image recognition and manipulation where we left off we were bringing up the images that we were going to consider and talking about thresholding and why we would want to and showing you the images that we're going to apply a threshold to we did begin building our threshold and now we're actually going to continue building it and now what I think I'm going to do is for now we'll just comment this out with three little quotes and at the end of it hit three quotes and down here uh we're going to pass [Music] through threshold and the image that we're going to want to do is going to be I guess we'll start with uh ir3 uh so we'll go ahead and save that run it and you should see the or the uh the text for the first pixel and the second pixel is basically we've got a 5-second timer on this so these are the pixel valuations right and this was that one that was really hard to read so anyway I'm going to stop it now now that we understand where we are I'm going to go ahead and we don't need to be printing out every pixel instead what we're going to do is we're going to average each of the pixels in there so we're going to make our averaging function so if you've watched some of my other videos I use this a lot but uh what we're going to use is reduce then we're going to make a Lambda uh x y and it's going to be x + y and we're going to use each piics and we don't really care about Alpha so we're just going to do to three we could go all the way to four but then we might adjust Alpha I don't want to be dealing with Alpha right now U maybe in another video anyway uh so that's the beginning there and then we want to divide by the length of each pics uh to three oops to three um or you could just divide by three really since it's only going to have three numbers in it but uh just to keep the average function working out for us anyway so that's that and then what we want to do is go balance wow it's learn to spell balance balance array. append and what do we want to append we want to append average num to that balance array done next once we've done that for every pixel right that's going to average like the average number in the pixel so how bright or how dark is each you know red green blue of that pixel now and then we append that to the balance array now we want to know like what is the average right of that so we'll say balance wow I did it again balance equals reduce and then again and um and in fact let's just copy and paste this save ourselves some time paste and now the only thing we need to change is instead of each picks to three we do balance array and oops balance array so that'll give us an average of uh the balance then what we're going to do is for uh each and we'll do the same thing we did before for each row in uh um new array so we're going to start changing up this new array because again like a numpy array is immutable you're not allowed to really mess with it so we're going to mess with it uh this way uh for each row in new array and then for each pics in each row then we're going to say um if the average because again you don't want to go by pixel by or or you don't want to do like red by Green by Blue you want to do like a combination of them and that will tell you cuz again right if if you remember what I was explaining before since it's an additive if you have like the full red and the full green that will make it even brighter than just a full red right so if you have full red full green and full blue now you've got a bright white right and if you just have full red and full green you've got something else and so on and so um so we don't really want to know per pixel and like change like the red the green the blue per pixel NE necessarily to threshold uh some thresholding operations you do and it just really depends what your goal is but in in our case we're just trying to make black and white right so we don't really want to do that so we're going to just say uh if reduce and we'll take we'll take this function right here this is exactly what we want if paste if that is greater than the balance right so if this one is more than the previous one what we want to do is each picks or if it's more than not previous one if it's more than the average right then that means this is one of our lighter colors right so then we're let's just convert everything so we'll say each Pi zero uh so like the first part of that pixel right so the red equals 255 and then we're just going to copy and paste this a few more times paste paste paste so each picks and then one two three obviously this is Alpha but we'll edit that one as well so all these values are going to be 255 and that will make us a nice pretty white else so if that's not the case then what do we want to do well we want to give everything a zero value besides the alpha so we'll just copy this come down paste and then 0 0 0 and we leave Alpha and then at the end of this uh function what we want it to do is we could print it out right but actually what we want to do is what's called return and so that actually like makes it act like it's like a math function right so if you do 1 plus one that's going to return something so we want to return new array but it just so happens this will end up modifying the array anyway so it's not we probably didn't really need to return anything but we will so um coming back down here now let's do uh we'll uncomment everything out and hopefully we didn't screw anything up and so we're going to let's first let's look at it one last time without the threshold so we'll run this and we get this and we're going to modify this image so you remember what it looks like right now very hard to look at and see but if we now call threshold for ir3 and run it and actually I think ir3 let me look let's look at this one5 yeah so we want ir3 anyway save that run that now that we're running the threshold and we got balance arrays not defined let's go back up here and look at it one last time oh man I so badly want to call it blance okay try again still remember what the other one looked like dude I did it again how many blance arrays did I do in here what I's see reduce blance I did it on all of them let's see here balance why didn't anybody tell me this okay hopefully we don't see that again that's pretty embarrassing anyway okay so now you can see that um it's been thresholded right so we've got white white and black and it looks nice and uh thresholded thresh held so now uh we could do it the same thing we could run threshold for ir2 as well so let's do that r that and so now we've got threshold threshold threshold and now let's threshold this sentex image and see how that looks so let's just copy paste and we'll threshold oops r four save that run it and now we can see what I was talking about as far as blur is concerned like the syntax image uh is a little like faded on the edges like it's it's not like like perfect right edges right got no threshold almost and uh so anyway this one you can see that you know the threshold held worked pretty well but again let's bring up just the regular uh image here you can see how it kind of like it's not like a solid here and it kind of Fades to White like it's not like I don't know a really sharp picture so we could bring up the other one too so here it is so you can see like at the tips the pixels slowly Fade to a white right so it just makes the picture look a little softer to the eyeballs and but anyway on a thresholding operation that can kind of throw it for a loop a little bit um so you get kind of like these little jaggedy edges but in the case of this image here you can see it's actually got um quite like 70 70 pixels so it's like a 70 by a 70 so if you were comparing this image to another one these little chunks that are almost missing wouldn't make a huge difference as far as comparing uh pattern to pattern pattern um or image array to image array that wouldn't make as big of a difference as it would here if these were like really jaggedy or whatnot so anyways uh that's the basics of our threshold and so if you have different colored images and stuff you would want to use threshold to uh basically make everything on the same playing field you know so you're dealing with the same exact numbers when you're considering so that's going to conclude tutorial number 7 as always thanks for watching thanks for your support your subscriptions and until next time so