1
00:00:05,972 --> 00:00:06,806
Hello, everyone.

2
00:00:06,806 --> 00:00:10,643
I would like to introduce you
to the next week's full podcast,

3
00:00:11,077 --> 00:00:14,180
which will feature a spotlight
on animal tracking.

4
00:00:14,481 --> 00:00:18,852
Here in studio, I have Shivang and Lianjie.
Hello, Shivang! Hello Lianjie!

5
00:00:19,185 --> 00:00:22,188
And remotely, we have Alex in Africa.

6
00:00:23,023 --> 00:00:28,328
Alex, can you give us a quick summary
what you will be talking about?

7
00:00:28,661 --> 00:00:31,664
I'm just very excited by this world.

8
00:00:31,664 --> 00:00:35,502
First project to create an app
that allows us

9
00:00:35,502 --> 00:00:39,939
to digitize the ancient craft of animal
tracking for the first time.

10
00:00:40,006 --> 00:00:45,545
It's going to allow people who enjoy
nature to go, to go into wilderness areas

11
00:00:45,845 --> 00:00:48,815
and be able to make sense of all the marks
on the ground

12
00:00:48,815 --> 00:00:51,818
through this,
this cutting edge technology with HPE.

13
00:00:52,919 --> 00:00:55,789
And how did you accomplish this?

14
00:00:55,789 --> 00:00:58,558
So in our system,
we use machine learning technology

15
00:00:58,558 --> 00:01:02,862
to preprocess
the image, including, removing the noise

16
00:01:03,129 --> 00:01:06,633
and the segmenting images
to identify all the footprints.

17
00:01:06,766 --> 00:01:10,870
And if the images are already labeled,
then we use them as training data

18
00:01:10,870 --> 00:01:11,504
in the future.

19
00:01:11,504 --> 00:01:14,808
And if they are not, we're going to use
them, store them as a backup.

20
00:01:15,041 --> 00:01:17,777
And then we use the machine
learning, technology.

21
00:01:17,777 --> 00:01:20,780
We also use them
to infer the species of the footprints.

22
00:01:20,947 --> 00:01:26,252
And then we use a, deep convolutional
neural network science to do that.

23
00:01:26,519 --> 00:01:30,623
and we pre-trained a neural network
on large public datasets.

24
00:01:30,790 --> 00:01:35,462
And then we also fine tune them using the
footprints Alex has already collected.

25
00:01:35,862 --> 00:01:36,463
Shivang.

26
00:01:36,463 --> 00:01:38,465
This seems extremely complicated.

27
00:01:38,465 --> 00:01:40,900
Have you simplified it
somehow? Yeah, I'll talk about.

28
00:01:40,900 --> 00:01:42,836
We will talk about that in the podcast
for sure.

29
00:01:42,836 --> 00:01:46,873
a key component of simplifying
it is the system that we build around

30
00:01:47,273 --> 00:01:50,343
the whole, algorithm
that Lianjie mentioned, right?

31
00:01:50,343 --> 00:01:52,078
Where we host the algorithm.

32
00:01:52,078 --> 00:01:53,379
Where are the images coming from?

33
00:01:53,379 --> 00:01:56,850
So I'll talk about the apps that are there
and available for Android now,

34
00:01:57,150 --> 00:01:59,886
and the web apps
and the backend system that we have.

35
00:01:59,886 --> 00:02:00,320
Thank you.

36
00:02:00,320 --> 00:02:01,855
Gentlemen, please join us.

37
00:02:01,855 --> 00:02:04,457
extremely interesting topic.

38
00:02:04,457 --> 00:02:09,062
A great benefit to humanity preserving
animal tracking and the art, thereof.