1
00:00:01,840 --> 00:00:09,790
Next is the concept of straight the first neuron in the conversion layer was looking at these sets of

2
00:00:10,030 --> 00:00:12,210
pixels.

3
00:00:12,330 --> 00:00:18,990
Now the next neuron in the conversion layer will have a slightly shifted receptive field or the window

4
00:00:20,460 --> 00:00:30,660
this shift in the view from one neuron to another is called the Straight so if the next neuron is looking

5
00:00:30,660 --> 00:00:40,470
at these actual pixels you can see that this window which is for the next neuron is shifted to pixels

6
00:00:40,590 --> 00:00:41,190
to the right.

7
00:00:43,050 --> 00:00:51,210
So in this scenario we see that this window has a straight of to each neuron on the convolution layer

8
00:00:51,600 --> 00:00:56,870
will have a window shifted to pixels to the right.

9
00:00:56,880 --> 00:01:05,280
So if we continue this the third neuron will be looking at a set of quantified pixels starting from

10
00:01:05,280 --> 00:01:05,770
here.

11
00:01:07,800 --> 00:01:16,830
So this will be the field of view for the third neuron and so on similarly a straight or forward means

12
00:01:16,830 --> 00:01:19,990
that you are taking a jump of four pixels.

13
00:01:19,990 --> 00:01:27,520
When you look at the next neuron to the second neuron will straight away look at these sets of pixels

14
00:01:28,000 --> 00:01:33,970
which is four pixels shifted towards delayed.

15
00:01:33,990 --> 00:01:41,100
Now if you use a small straight then there would be a lot of overlap between these two deceptive feet

16
00:01:42,250 --> 00:01:50,020
so if you look at the state of two you can see that this three by five rectangle is common between the

17
00:01:50,620 --> 00:01:57,150
receptive field of neuron 1 and deceptive field of neuron 2 2.

18
00:01:58,110 --> 00:02:01,970
We have 15 pixels common for neuron 1 and you don't do.

19
00:02:02,640 --> 00:02:09,270
But if you have a straightforward in that scenario only this single line is common for neuron 1 and

20
00:02:09,270 --> 00:02:12,180
neuron do so only 5 pixels are common.

21
00:02:12,360 --> 00:02:19,740
So there is small overlap if this trade is large and there is larger overlap if straight is small.

22
00:02:20,460 --> 00:02:28,980
Also if this trade is large since the overlap will be less therefore fewer neurons will be required

23
00:02:29,100 --> 00:02:38,880
in the upper layer so if we have a trade off for many neurons in the upper layer will be less than if

24
00:02:38,880 --> 00:02:49,250
we had a straight of to so the trade will determine the size of a player and the amount of overlap in

25
00:02:49,250 --> 00:02:50,240
the receptive field.