Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:00,930 --> 00:00:03,400 Hello and welcome back to the course on computer vision. 2 00:00:03,490 --> 00:00:06,040 And today we're going to talk about cascading. 3 00:00:06,100 --> 00:00:12,790 So previously we discussed about the show on classifiers and how they're constructed from Week classifiers 4 00:00:12,790 --> 00:00:13,860 and put them together. 5 00:00:14,140 --> 00:00:19,780 And now we're going to talk about another hack that the burial of Jones algorithm applies in order to 6 00:00:19,780 --> 00:00:20,950 speed up the process. 7 00:00:21,220 --> 00:00:25,420 And the hack is called Cascadian schematically it looks like this. 8 00:00:25,480 --> 00:00:28,170 We take a window that we're evaluating on our. 9 00:00:28,170 --> 00:00:37,150 So that's a window that is sliding across our lives and we look for the first feature in all or list 10 00:00:37,150 --> 00:00:37,540 of features. 11 00:00:37,540 --> 00:00:43,210 So for the most important feature and then we see if it's present on the image or in that same window 12 00:00:43,210 --> 00:00:51,040 if it's present in some window or not if it's not present in the same window then we reject that window 13 00:00:51,040 --> 00:00:53,500 completely and we don't even evaluate the remainder of the picture. 14 00:00:53,500 --> 00:00:59,410 So if for instance this is the most water feature and we find out that it's not present in some window 15 00:00:59,620 --> 00:01:02,400 then we're not even going to look at these. 16 00:01:03,070 --> 00:01:10,510 Then if you say the way to think about it is if if there is no nos an image then it's just not going 17 00:01:10,510 --> 00:01:16,210 to be a face like a frontal face has to have a nose therefore what's the point of looking for eyes and 18 00:01:16,210 --> 00:01:19,540 cheeks and eyebrows and so on if we're you know that there's no nodes. 19 00:01:19,840 --> 00:01:27,070 Then if this feature is in the presence of if there is a nose then it is time to walk for the second 20 00:01:27,070 --> 00:01:27,540 feature. 21 00:01:27,550 --> 00:01:31,880 Now we're going to evaluate if the second features in that sampling them if it's not then we reject 22 00:01:31,890 --> 00:01:38,620 the subway for instance if there's a nose but then whatever this is there's this looks thing an eyebrow 23 00:01:38,650 --> 00:01:42,320 but it's upside down so maybe there's no Water-Lily. 24 00:01:42,560 --> 00:01:47,500 Whatever this feature represents whatever the algorithm was looking for there. 25 00:01:47,800 --> 00:01:53,650 It's not present then we're going to now reject the reject image because it has a nose and doesn't have 26 00:01:53,650 --> 00:01:55,310 a lip so on your face. 27 00:01:55,330 --> 00:02:01,130 So then if certain value the third feature if it is present if that should also happen then and without 28 00:02:01,250 --> 00:02:02,090 talked. 29 00:02:02,650 --> 00:02:03,340 And so on. 30 00:02:03,340 --> 00:02:08,250 So if servies is not present then we reject and move on. 31 00:02:08,710 --> 00:02:13,640 So that's the simplistic diagrammatic kind of explanation. 32 00:02:13,660 --> 00:02:15,450 In reality it's a bit more complex. 33 00:02:15,530 --> 00:02:22,660 In reality of course it's very risky to rely on just one feature because maybe it might not be. 34 00:02:22,900 --> 00:02:28,240 Maybe the shade is in such a way or the lighting is different or maybe just so for some reason wasn't 35 00:02:28,240 --> 00:02:34,660 detected as can be lots of different circumstance so what it actually does is in this first step it 36 00:02:34,660 --> 00:02:39,110 doesn't look at the one feature it looks like the top five features with the top 12 inches. 37 00:02:39,260 --> 00:02:44,650 Well let's say let's say Tokai for argument's sake it looks at top five features and then decides based 38 00:02:44,650 --> 00:02:51,220 on them so if if they're not present if for instance none of them are present then or some portion of 39 00:02:51,220 --> 00:03:00,820 them are present like the details the details are not important right now what is important is the whole 40 00:03:00,820 --> 00:03:07,930 concept that regardless if you find identifies based on the first five features it takes one of two 41 00:03:07,930 --> 00:03:14,530 or three or four or five and it's based on them that this window this some Windu is pointless had to 42 00:03:14,530 --> 00:03:20,590 keep looking at it's not a face it won't continue then here it will look at the next 12 here or look 43 00:03:20,590 --> 00:03:22,310 at the next 25 or something. 44 00:03:22,330 --> 00:03:27,250 So every time because the features are becoming less and less important less and less prominent is taking 45 00:03:27,250 --> 00:03:28,660 more and more of these features. 46 00:03:28,870 --> 00:03:36,280 But nevertheless is basing its decision on whether or not to continue going forward with this on the 47 00:03:37,690 --> 00:03:39,940 at the at that point in time all the features it's looking at. 48 00:03:39,940 --> 00:03:46,720 And if it is not happy with one and seeing things that it might now like if it decides that over here 49 00:03:46,720 --> 00:03:52,690 that right now this should be the these features and all this many of these features that are not there. 50 00:03:52,690 --> 00:03:53,340 This is not a face. 51 00:03:53,350 --> 00:03:55,020 I'm not even going to evaluate the rest. 52 00:03:55,150 --> 00:03:56,550 And that is called cascading. 53 00:03:56,590 --> 00:04:00,170 It helps really speed up the process. 54 00:04:00,190 --> 00:04:04,450 So in visual terms let's say that these are all features that we identify. 55 00:04:04,540 --> 00:04:09,510 Again we'll go back to the simple example of just getting them one by one rather than in batches. 56 00:04:09,730 --> 00:04:13,490 And so what we're going to look at is we're going to look at the subgroup over here. 57 00:04:13,550 --> 00:04:18,090 And so let's say we look at this feature and we can see that it is present. 58 00:04:18,100 --> 00:04:19,900 That's the eyebrow over here. 59 00:04:19,900 --> 00:04:20,740 Dark and bright. 60 00:04:20,740 --> 00:04:23,230 So this will pass so that that passes. 61 00:04:23,230 --> 00:04:24,580 OK so it goes from here to here. 62 00:04:24,580 --> 00:04:25,140 Good. 63 00:04:25,150 --> 00:04:27,010 Second Future looks for this. 64 00:04:27,010 --> 00:04:28,310 You can see that in the eye. 65 00:04:28,360 --> 00:04:35,260 OK it passes then goes to this and maybe in this case it actually can see that. 66 00:04:35,260 --> 00:04:45,040 So this is technically this feature is reserved or was originally planned for the lips so light black 67 00:04:45,310 --> 00:04:51,760 lights a light dark light and it would that would have identify liberties we've got an intuition that 68 00:04:51,970 --> 00:04:58,210 actually eyeball you can see the same thing you can see light dark and light so it can mistakenly actually 69 00:04:58,780 --> 00:05:03,560 accept the this feature is here and would say oh this features here so we do have the lips and really 70 00:05:03,560 --> 00:05:04,190 we don't. 71 00:05:04,250 --> 00:05:10,110 But OK so even if that passes it gets to this future and it cannot identify this feature because this 72 00:05:10,110 --> 00:05:15,860 feature for instance like nowhere on this image you can see like light and dark until dark and you can 73 00:05:15,860 --> 00:05:20,350 see the light and dark of light but not the other way around the white and black like that. 74 00:05:20,390 --> 00:05:23,990 So it was actually for this part of the nose as far as I remember. 75 00:05:24,140 --> 00:05:28,730 So at this point it would at this point it would reject I would say this feature features are present 76 00:05:28,730 --> 00:05:30,530 so featuring them for the present. 77 00:05:30,680 --> 00:05:34,790 They reject the sub window completely and it wouldn't even evaluate this feature. 78 00:05:34,850 --> 00:05:39,440 And as you can imagine you can get kind of 100 features or how many features after that it would just 79 00:05:39,440 --> 00:05:41,080 not evaluate them any of them. 80 00:05:41,210 --> 00:05:46,010 And even though evaluation takes like split seconds for each one but when you add them up because there's 81 00:05:46,010 --> 00:05:52,590 so many of them it takes time and plus you have to go through this image lots of time with some windows. 82 00:05:52,970 --> 00:05:59,480 And so it saves not just time devaluing the Swan But lots of features all the features that come afterwards. 83 00:05:59,480 --> 00:06:02,000 And it just rejects this whole sub window and moves on. 84 00:06:02,000 --> 00:06:05,780 And that really speeds up the process and that's called cascading. 85 00:06:05,780 --> 00:06:08,030 So on that note we're going to wrap up today's tutorial. 86 00:06:08,030 --> 00:06:11,960 I hope you enjoyed it and I look forward to seeing you next time. 87 00:06:11,960 --> 00:06:13,670 And until then enjoy computer vision. 9787