1
00:00:00,000 --> 00:00:04,066
[MUSIC]

2
00:00:04,066 --> 00:00:05,597
So in this deep learning module,

3
00:00:05,597 --> 00:00:09,320
we talked about two really cool things
that you can do with deep learning.

4
00:00:09,320 --> 00:00:15,150
One is classify images, and
the other one is retrieve images.

5
00:00:15,150 --> 00:00:17,510
And in both of them,
we talked about deep features.

6
00:00:17,510 --> 00:00:20,490
So let's just go ahead and
build a notebook for

7
00:00:20,490 --> 00:00:23,390
image classifications using deep features.

8
00:00:23,390 --> 00:00:26,470
So, let's give this notebook a name.

9
00:00:26,470 --> 00:00:32,829
So, this is going to be
Image Classification

10
00:00:32,829 --> 00:00:37,850
with Deep Features.

11
00:00:37,850 --> 00:00:41,490
And I'm going to hide the header and

12
00:00:41,490 --> 00:00:44,730
the toolbar to get a little
bit more space, as usual.

13
00:00:44,730 --> 00:00:46,980
And we'll fire up.

14
00:00:46,980 --> 00:00:51,945
Actually I just did a little typo here so

15
00:00:51,945 --> 00:00:57,158
we will, Fire up GraphLabCreates.

16
00:00:57,158 --> 00:01:02,786
So let's import GraphLab, So here we go.

17
00:01:02,786 --> 00:01:09,309
So the first thing that we're going to do
is load a common image analysis dataset.

18
00:01:19,689 --> 00:01:23,660
So this is a benchmark
dataset called a SIFR10.

19
00:01:26,470 --> 00:01:30,220
We select a subset of those where
we just have four categories

20
00:01:30,220 --> 00:01:32,460
of objects we're trying to recognize.

21
00:01:32,460 --> 00:01:36,430
So I'm going to load the data set, and

22
00:01:36,430 --> 00:01:40,540
this data set has already been split
into train data and test data,

23
00:01:40,540 --> 00:01:45,660
because in computer vision, often that
split happens so every researcher uses

24
00:01:45,660 --> 00:01:50,580
the same training set and the same test
set, so the papers become more comparable.

25
00:01:50,580 --> 00:01:58,150
So, I'm just gonna load an S
frame that contains the data set.

26
00:01:58,150 --> 00:02:02,790
So, graphlab.SFrame and for the training

27
00:02:04,000 --> 00:02:09,046
data, there's a directory
here called image_train_data.

28
00:02:09,046 --> 00:02:16,110
And then for the test data, so
let's call that image_test.

29
00:02:16,110 --> 00:02:21,109
We have another SFrame.

30
00:02:21,109 --> 00:02:25,681
Code image_test_data.

31
00:02:25,681 --> 00:02:29,830
Here we go, so
we are loading this two data sets,

32
00:02:29,830 --> 00:02:33,460
and let's start by exploring
the data a little bit.

33
00:02:33,460 --> 00:02:39,320
So, exploring the image data.

34
00:02:42,300 --> 00:02:45,390
So, what we're gonna do is use canvas.

35
00:02:45,390 --> 00:02:48,700
So, I'm gonna point canvas
to show it on the notebook,

36
00:02:48,700 --> 00:02:50,550
as we have been doing
on several notebooks.

37
00:02:50,550 --> 00:02:53,240
So, it becomes all one thing
that we can take home.

38
00:02:53,240 --> 00:02:59,145
So, I'm gonna set graphlab.canvas and

39
00:02:59,145 --> 00:03:07,090
I'm going to set the target
to be the ipython notebook.

40
00:03:07,090 --> 00:03:10,510
So it's going to print anything on
the iPython notebook that we do.

41
00:03:10,510 --> 00:03:15,050
And let's start by taking
a look at the string data.

42
00:03:15,050 --> 00:03:17,350
So image string, and
here's a few columns but

43
00:03:17,350 --> 00:03:20,898
the column that we really care
about here is the image column.

44
00:03:20,898 --> 00:03:23,110
I am just going to type .show on that and

45
00:03:23,110 --> 00:03:26,960
we are going to take a quick
look at that image column.

46
00:03:26,960 --> 00:03:31,039
And these images are not too large,
so I'm temporarily going to increase

47
00:03:32,320 --> 00:03:37,070
the font here so
we can take a look at the images.

48
00:03:37,070 --> 00:03:42,070
And so if you look at it, there's a bird,
there's some cats, there's some dogs and

49
00:03:42,070 --> 00:03:42,828
some cars.

50
00:03:42,828 --> 00:03:49,433
So this data set is all about cats,
dogs, cars, and birds.

51
00:03:49,433 --> 00:03:53,659
[MUSIC]