Data labelling in AI Center

Help AI Center
,
1

How we can create a data labeling session in AI Center. and how to label data. What is the function of all the optoins.

2

Hi @Aditi_Jha

Check the below docs:

Regards

3

@Aditi_Jha

It a very broad question. Better register is academy and you would find videos related to ai center which would help

Cheers

1 Like
4

Just one doubt, minimum how many documents do we need to train in the data labelling session? And once we train all the documents, what is the next step? Do we need to export the trained document and then upload it to data sets?

5

@Aditi_Jha

Labelling should be done for minimum of 5 documents per type…the more you do it is more better

Are you training in studio? If you are creating the labelling session in ai center then next step would be to choose a model and then create a pipeline selecting the dataset and the mode you choose to create the skill

Cheers

6

I am training in AI Center. So which data sets I will be selecting, the ones that I trained in data labelling, right?

7

@Aditi_Jha

Yes you would select the same for training pipeline…and you would select the lbelling session and the respective model you would want to train on

Cheers

8

Suppose I have labelled data in a set of documents, then how it will be used further. ? I should be uploading these trained documents in datasets, right?

9

@Aditi_Jha

After uploading documents only you would do the lbelling in the ai center…why would you upload again?

Cheers

10

So the documents in datasets will know that they are trained? And what if we do not label data and we use them directly in pipeline?

11

@Aditi_Jha

In the pipeline you would select the labelled set only…else there is no point of labelling…

If the respective model is already a pretrained one then you dont need to train again but rather consume

So this is how it goes

  1. Upload files to dataset
  2. Label the documents and then export which will created the json in dataset
  3. select a package
  4. Create a pipeline and select the labelled documents as dataset(which is exported after labelling) and package selected in 3 and train
  5. Using the trained pipeline version create a skill

Cheers

12

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.