1
00:00:00,870 --> 00:00:01,180
Hello.

2
00:00:01,230 --> 00:00:01,800
Welcome back.

3
00:00:01,920 --> 00:00:07,010
Let's analyze a complete convolution or your network and this arrangement over here.

4
00:00:07,020 --> 00:00:16,260
We take an input image of size fetch a 9 by fetch a 9 by 3 and we confirm it with 10 tests of size through

5
00:00:16,260 --> 00:00:21,800
by 3 using a pattern of 0 and astride of 1.

6
00:00:21,870 --> 00:00:30,420
As we have written over here this gives us an output of thirty seven by thirty seven by ten.

7
00:00:30,420 --> 00:00:40,890
We convert this output with twenty four tests of size five by five applying astride of two and pattern

8
00:00:40,980 --> 00:00:41,910
of zero.

9
00:00:41,910 --> 00:00:48,240
This will give us seventeen by seventeen by twenty outputs as we have over here.

10
00:00:48,240 --> 00:00:55,710
We take this new output and control it with 40 50 tests of size 5 by 5.

11
00:00:56,010 --> 00:01:05,070
Applying astride of two and a pattern of zero this will give us 7 by 7 by 40 as we see over here.

12
00:01:05,070 --> 00:01:12,350
Finally we take this three dimensional output and unroll it into a single vector.

13
00:01:12,360 --> 00:01:19,620
This gives us one thousand nine hundred and sixty long column vector seven multiplied by seven multiplied

14
00:01:19,620 --> 00:01:20,460
by 40.

15
00:01:20,460 --> 00:01:24,210
Of course one thousand nine hundred and sixty.

16
00:01:24,210 --> 00:01:30,150
This is a short summary of convolution or neural network in convolution or neural networks.

17
00:01:30,150 --> 00:01:32,410
We have three types of letters.

18
00:01:32,460 --> 00:01:34,100
We have the convolution on there.

19
00:01:34,230 --> 00:01:40,950
We have the constitutional layer that Portlandia and d formerly connected layer before we examine an

20
00:01:40,950 --> 00:01:44,090
example depicting these three types of layers.

21
00:01:44,100 --> 00:01:47,160
Let's first talk about the polling data in the next lesson.