Form Recognizer Labeling Tool - working with remote team - best practice

Question

Form Recognizer Labeling Tool - working with remote team - best practice

Naqqash 21

Hi there,

I am working with a remote team and want to give access to the same labeling tool project(s) I am working on. Currently, I am using a docker container locally for labelling and custom models but I could not find enough information on what is the recommended option about working on the labeling tool across teams based on different regions.

One possibility that comes to my mind is to deploy the same container on Azure portal and give access to the relevant team member?

Was wondering if there is another suggestion.

Regards,
Naqqash

Accepted answer

0 additional answers

Your answer

Answer 1

romungi-MSFT 48,911 Microsoft Employee Moderator

@Naqqash Since you are using a blob container with your docker container hosted labeling tool you can share the same blob container SAS URI to your team and request them to add the same as a new connection with either the hosted azure version of the labeling tool with this link.
https://fott-preview.azurewebsites.net/connections/create

You can also add a new connection with the docker container using your URI https://<your_host>/connections/create

Once this is done you should be able to open the projects under this container using Open cloud project and select the connection that is created.

Another simple way to achieve the same is to share your project from the top right corner.

Share this token to your team and request then to open cloud project option to paste the key from their version of labeling tool.

I would recommend to sue the same version of the labeling tool to avoid any issues with compatibility.

Naqqash 21 Reputation points

2021-08-23T16:23:40.983+00:00

Thank you @romungi-MSFT .

I am using now the second option from your answer. From the scalability aspect still does not seem to be the best option but should work for now.

Since, I am on the topic of scalability what are the general recommendations for scaling this. Surely, this approach can not work in the long term when I will be dealing with 100+ different types.

I have found this useful resources to automate the labelling to a certain extent but are there any suggestions as "best practices".

https://github.com/microsoft/knowledge-extraction-recipes-forms/tree/master/Training/Auto_Labelling/basic_implementation

This link has been also very valuable;
https://github.com/microsoft/knowledge-extraction-recipes-forms

Another question not entirely related to this is; in my Azure portal, where can I see my model e.g. If I want to delete existing models, I don't think I can do anything from the compose tab other than selection for analysis.

Regards,
Naqqash
romungi-MSFT 48,911 Reputation points Microsoft Employee Moderator

2021-08-24T06:29:58.377+00:00

@Naqqash Do you mean sharing the labeling tool or project for 100+ users? I think the labeling tool is designed to build and train models which can then be used via the SDKs or the REST API which are scalable. Sharing for labeling tool project is designed for limited users or developers to create a model that performs to the extent desired. You can raise a github issue with the labeling tool github repo with more details of the scalability aspect so the team could consider enhancements in the future release.

To delete models you can use the REST API or any SDK that is available based on the language that works for you from the references section of FR documentation.
Naqqash 21 Reputation points

2021-08-24T08:50:10.193+00:00

Thank you for your support.

Regards,
Naqqash

Share via

Form Recognizer Labeling Tool - working with remote team - best practice

0 additional answers

Your answer