1
00:00:00,670 --> 00:00:07,330
In this video, we are going to make a convolutional neural network model and we will train our model

2
00:00:07,600 --> 00:00:10,510
to predict the fashion object in the images.

3
00:00:11,020 --> 00:00:13,970
So we'll be using fashion amnesty doggin.

4
00:00:15,940 --> 00:00:20,590
We have created Ian and modelled on this dataset in the previous section.

5
00:00:21,550 --> 00:00:24,100
Now we are going to create a CNN model.

6
00:00:25,870 --> 00:00:33,700
The architecture of the model is going to be this last time when we created in and model this right

7
00:00:33,700 --> 00:00:35,330
and take part while our model.

8
00:00:36,100 --> 00:00:37,220
We flatten the layer.

9
00:00:37,990 --> 00:00:40,810
We put two dense layers, which were the hidden layers.

10
00:00:40,990 --> 00:00:43,080
The first head and led had 300 neurons.

11
00:00:43,390 --> 00:00:44,680
Second had 100 neurons.

12
00:00:44,770 --> 00:00:50,800
And lastly, it had then you not know to make CNN model.

13
00:00:51,270 --> 00:00:54,790
We are going to add two new two new layers.

14
00:00:55,240 --> 00:00:57,490
The First Lady will be convolutional layer.

15
00:00:59,380 --> 00:01:02,830
This layer will take out 32 features.

16
00:01:03,590 --> 00:01:11,230
So we will have 32 different windows and therefore it will give us 32 features from the input images

17
00:01:13,270 --> 00:01:21,310
on this convolutional layer will put up pulling it, pulling Lyd will aggregate the output of neurons

18
00:01:21,340 --> 00:01:26,800
in the congressional layer and therefore reduce the computations in the network.

19
00:01:27,790 --> 00:01:32,620
So this lead is added to make the network computationally more efficient.

20
00:01:34,090 --> 00:01:36,940
So these are the only two layers we are adding for now.

21
00:01:38,680 --> 00:01:43,990
The actual convolutional models have a stack of convolutional and pooling layers.

22
00:01:44,770 --> 00:01:48,510
But for now, we just add one kostner and one pulling.

23
00:01:50,430 --> 00:01:56,950
The output of pulling layer will then be input into the normal artificial neural network model that

24
00:01:56,950 --> 00:01:57,850
we created lasting.

25
00:01:59,950 --> 00:02:03,210
So this is the architecture on which we will be working on this thing.