Skip to content

SMGT Ground Truth Combine Bounding Box and Polygon

0

For example I have 10 labels, out of which I want 8 to be bounding box and rest 2 to polygon. Is there a way that I can add option for that?

Currently the best I have is using 2 tabs one have crowd-polygon element and other with crowd-bounding-box and with respective labels. But the issue is both them have submit buttons which can be confusing. If this is the best solution, is there a way to hide these submit buttons and add a single button in crowd-form

1 Answer
0

Amazon SageMaker Ground Truth doesn't directly support mixing different annotation types (like bounding boxes and polygons) within a single labeling interface out of the box. Your approach of using two separate elements is on the right track.

When creating a custom labeling interface in SageMaker Ground Truth, you're working with Crowd HTML Elements. The solution you've described using separate tabs for crowd-polygon and crowd-bounding-box elements is a workable approach, but I understand the concern about having multiple submit buttons.

To create a single submit button, you would need to customize your HTML template. Since you're already using a custom template with crowd-form, you can modify it to hide the default submit buttons from the individual elements and create a single submit button at the form level.

Here's how you might approach this:

  1. Use CSS to hide the default submit buttons in the crowd-polygon and crowd-bounding-box elements
  2. Add a single submit button within your crowd-form element
  3. Use JavaScript to handle the submission of both annotation types when the single button is clicked

This would require more advanced customization of the HTML template, but it would provide a cleaner user experience with a single submit button.

SageMaker Ground Truth supports various built-in task types including bounding box, semantic segmentation, and polygon annotations, but combining them requires this type of custom solution.
Sources
Automate Labeling for Intelligent Document Processing with Inawisdom and Amazon SageMaker Ground Truth | AWS Partner Network (APN) Blog
Label verification and adjustment - Amazon SageMaker AI

answered 10 months ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.