1
00:00:00,000 --> 00:00:02,190
What does it take to download

2
00:00:02,190 --> 00:00:04,050
a public dataset
off the Internet,

3
00:00:04,050 --> 00:00:05,175
like cats verses dogs,

4
00:00:05,175 --> 00:00:07,920
and get a neural network
to work on it?

5
00:00:07,920 --> 00:00:09,990
Data is messy, sometimes

6
00:00:09,990 --> 00:00:12,060
you find surprising
things like pictures of

7
00:00:12,060 --> 00:00:13,320
people holding cats or

8
00:00:13,320 --> 00:00:16,590
multiple cats or
surprising things in data.

9
00:00:16,590 --> 00:00:19,019
In this week, you get to practice

10
00:00:19,019 --> 00:00:22,545
with using TensorFlow to deal
with all of these issues.

11
00:00:22,545 --> 00:00:24,450
Yeah, and it's like,
so even for example,

12
00:00:24,450 --> 00:00:25,530
you might have
some files that are

13
00:00:25,530 --> 00:00:28,290
zero length and they could
be corrupt as a results.

14
00:00:28,290 --> 00:00:30,300
So it's like using
your Python skills,

15
00:00:30,300 --> 00:00:31,530
using your TensorFlow skills

16
00:00:31,530 --> 00:00:33,105
to be able to filter them out.

17
00:00:33,105 --> 00:00:34,830
Building a convolutional net

18
00:00:34,830 --> 00:00:36,330
to be able to spot things
like you mentioned,

19
00:00:36,330 --> 00:00:37,680
a person holding it up.

20
00:00:37,680 --> 00:00:39,350
So that's some of
the things we'll

21
00:00:39,350 --> 00:00:40,870
do this week, is by using,

22
00:00:40,870 --> 00:00:42,890
and it's still
a very clean dataset

23
00:00:42,890 --> 00:00:44,600
that we're using with
cats versus dogs,

24
00:00:44,600 --> 00:00:46,200
but you're going to hit
some of those issues.

25
00:00:46,200 --> 00:00:49,310
I think you'll learn
the skills that you need to be

26
00:00:49,310 --> 00:00:50,630
able to deal with

27
00:00:50,630 --> 00:00:53,020
other datasets that may not
be as clean as this one.

28
00:00:53,020 --> 00:00:54,725
Yeah. Sometimes people think that

29
00:00:54,725 --> 00:00:56,540
AI is people like Lawrence and

30
00:00:56,540 --> 00:00:58,340
me sitting in front of

31
00:00:58,340 --> 00:01:00,685
a white board maybe a
zen garden outside,

32
00:01:00,685 --> 00:01:03,480
talking about
the future of humanity.

33
00:01:03,480 --> 00:01:05,930
The reality is, there's
a lot of data cleaning,

34
00:01:05,930 --> 00:01:07,940
and having great tools
to help with

35
00:01:07,940 --> 00:01:09,080
that data cleaning makes

36
00:01:09,080 --> 00:01:10,745
our workflow much more efficient.

37
00:01:10,745 --> 00:01:11,605
Definitely.

38
00:01:11,605 --> 00:01:12,840
So in this week,

39
00:01:12,840 --> 00:01:14,680
you get to practice all that,

40
00:01:14,680 --> 00:01:17,045
as well as train
a pretty cool neural network

41
00:01:17,045 --> 00:01:20,160
to classify cats versus
dogs. Please dive in.