1
00:00:00,750 --> 00:00:01,130
Hello.

2
00:00:01,140 --> 00:00:02,350
Welcome back.

3
00:00:02,370 --> 00:00:09,210
When dealing with our Richie B call colored images the dimension of each input image is three dimensional

4
00:00:09,300 --> 00:00:13,570
because the other three color Chanos red green and blue.

5
00:00:13,770 --> 00:00:19,920
Because of this the dimension of the filter can or must also be three dimensional one filter for each

6
00:00:19,920 --> 00:00:26,310
colour channel of the inputs image over here we can see a six by six by three inputs.

7
00:00:26,310 --> 00:00:30,650
Image converged with a three by three by three filter Kanno.

8
00:00:31,110 --> 00:00:34,040
The outputs shape will be four by four.

9
00:00:34,620 --> 00:00:42,210
So the first element of the outputs is the sum of the elements Y's product of the first twenty seven

10
00:00:42,210 --> 00:00:44,050
values from the input.

11
00:00:44,100 --> 00:00:52,770
Remember nine values from each channel and the twenty seven values from the filter the two turn seven

12
00:00:52,770 --> 00:00:59,340
values from the input and the twenty seven values from the filter gives us the first element of the

13
00:00:59,340 --> 00:01:00,770
output.

14
00:01:00,780 --> 00:01:03,780
After this we convert the whole image.

15
00:01:03,780 --> 00:01:13,530
In this process also instead of using a single three dimensional filter we can use multiple three dimensional

16
00:01:13,530 --> 00:01:15,570
filters as well.

17
00:01:15,570 --> 00:01:21,750
We could do this because we want each three dimensional filter to detect a different type of feature.

18
00:01:21,750 --> 00:01:28,770
For instance we want one filter can or two detect horizontal edges and another filter can or two detect

19
00:01:28,920 --> 00:01:30,730
vertical edges.

20
00:01:30,780 --> 00:01:37,350
In this instance the dimension of our output matrix will be four by four by two because we have two

21
00:01:37,440 --> 00:01:40,260
outputs as we can see in this arrangement.

22
00:01:40,290 --> 00:01:45,400
Over here we can generalize the equation of the output actually.

23
00:01:45,870 --> 00:01:53,790
We can say if and see this the number of channels of the input image and NC Prime is the number of channels

24
00:01:53,850 --> 00:02:05,580
of the film Kano one given an input image of size and by and by NC the output will be and plus two p

25
00:02:05,820 --> 00:02:19,650
minus f divided by s plus one by and plus two p minus f divided by s plus one by number of channels

26
00:02:19,650 --> 00:02:26,050
which is and primacy over here or NC prime number of channels of the fields.

27
00:02:26,050 --> 00:02:27,680
I can't all right.

28
00:02:28,080 --> 00:02:32,040
So that's all the risk for this lesson and I shall see you in the next lesson.