WEBVTT - generated by wenglor-media

1
00:00:00.440 --> 00:00:02.040
Now that our images are uploaded,

2
00:00:02.040 --> 00:00:04.400
let’s label them inside the AI Lab.

3
00:00:04.440 --> 00:00:06.600
To start, we first need to create classes.

4
00:00:06.840 --> 00:00:09.960
On the landing page, directly add the two classes:

5
00:00:10.280 --> 00:00:12.880
OK and NOK to your dataset.

6
00:00:13.440 --> 00:00:16.360
You can adjust the color coding for classes
at any time.

7
00:00:16.560 --> 00:00:19.840
We now have different ways of adding labels
to our images.

8
00:00:20.280 --> 00:00:22.080
First, we can click on an image

9
00:00:22.080 --> 00:00:24.080
to open it in full view.

10
00:00:24.160 --> 00:00:27.800
Then, use the number keys 1 or 2 to assign
a class —

11
00:00:27.800 --> 00:00:32.120
for example, “1” for OK and “2” for
NOK.

12
00:00:40.920 --> 00:00:43.480
We can also select multiple images at once

13
00:00:43.480 --> 00:00:45.480
and label them together.

14
00:00:53.280 --> 00:00:54.360
For batch labeling,

15
00:00:54.360 --> 00:00:56.560
you can filter images in your dataset —

16
00:00:56.560 --> 00:00:57.320
for example,

17
00:00:57.320 --> 00:01:00.000
by the tags you added earlier in uniVision
3.

18
00:01:00.000 --> 00:01:03.000
Let’s filter for all images tagged as “OK”,

19
00:01:05.000 --> 00:01:07.000
select them and assign a “Ground Truth”.

20
00:01:16.000 --> 00:01:19.000
If a “Ground Truth” was assigned by mistake,

21
00:01:19.000 --> 00:01:21.000
you can easily remove it again.

22
00:01:39.000 --> 00:01:43.000
Next, filter for all images that are not annotated
—

23
00:01:43.000 --> 00:01:45.000
these are the ones without any class yet.

24
00:01:47.000 --> 00:01:51.000
Select them all and assign the NOK class.

25
00:01:54.000 --> 00:01:55.000
And that’s it —

26
00:01:55.000 --> 00:01:59.000
all images in your dataset are now labeled
and ready for training.