Annotate

In

User Guide

3 mins

Read

Annotab Team

Bounding Box

Bounding box is a rectangular shape, drawn to mark objects of interest. Similar to any rectangle, a bounding box is defined by two points. Users click on the initial point and then drag it to the second point, effectively creating the bounding box. While bounding boxes are typically effective for determining object position in rectangular images, they may fall short when dealing with non-rectangular images. In such cases, alternative methods are required for accurate object detection.

In Annotab Studio, to segment with Bounding Box:

1. Click

2. Identify the object you want to annotate within the image.

3. Click on a point on your object.

4. Hold & drag the mouse to the desired point.

5. Release the mouse, and a bounding box will auto be created.

6. “Class” pop up will appear.

  1. You have 2 options:

    1. Search for the available classes from the drop down list or type in the name of the class.

    2. If the class is not available, the Class pop up will show additional options to create class.

8. Click "Save" to complete the process.




Polygon

Polygon provides more precise coverage by allowing an arbitrary number of points, but it can be challenging to draw and use for annotators. Drawing polygons accurately requires attention to detail, making it a complex process.

In Annotab Studio, to segment with Polygon:

1. Click

2. Identify the object you want to annotate within the image.

3. Begin by clicking on the starting point of the object boundary.

4. Continue clicking on additional points along the object's boundary, forming a series of connected lines.

5. Once you reach the final point, complete the polygon by connecting it to the starting point.

6. Ensure that the polygon encompasses the entire object accurately.

  1. “Class” pop up will appear.

  2. You have 2 options:

    1. Search for the available classes from the drop down list or type in the name of the class.

    2. If the class is not available, the Class pop up will show additional options to create class.

9. Click "Save" to complete the process.




Auto-Segment

Auto-Segment is developed by Annotab AI to help automatically segment items and create pixel-perfect polygon masks by leveraging deep learning.

In Annotab Studio, to segment with Auto-Segment:

1. Choose your preferred model. Currently, we offer 3 kinds of models: Segment Anything Model (SAM), Rabbit or your own model (Coming soon).

2. Click

3. Identify the object you want to annotate within the image.

4. Begin by drawing a rough bounding box around the object boundary. Leave a bit of padding around the object.

5. Auto-Segment will locate the most salient object in that boundary.

6. Click “Finish” to save your segmented object.

  1. “Class” pop up will appear.

  2. You have 2 options:

    1. Search for the available classes from the drop down list or type in the name of the class.

    2. If the class is not available, the Class pop up will show additional options to create class.

8. Click "Save" to complete the process.




What's next

Send your Image to Review

Once you've completed annotating your data, click \"Send to Review\" and your designated reviewer will be notified. Monitor the status of your annotated data under \"Review\" status in Dataset > Data.

/screenshot/

Review

As a reviewer, you can ensure that labelled data is free from errors & qualified to become ground truth by accepting or rejecting images. To start reviewing:

  1. Find images with this status

  2. Mark “Accept” or “Reject” depending on how you assess the label quality. Both accepted and rejected images will be moved forward to next steps in the workflow, depending on how you initially set them up.

  3. Marking “Issue” where you find defects that need to be resolved. Reviewer & annotators can view list of issue in the right side bar and choose to resolve or leave the issue. Unresolved issue will not affect the acceptance/rejection process.

/screenshot/

View Also

  • Analytics

  • Export Data

Updated

Nov 8, 2023

Found helpful?

Updated

Dec 20, 2023

© 2024 Annotab Pte. Ltd.

Found helpful?

Create a Project

Subtopic desc.

In

Content

Content

Read

Content

Updated

Dec 20, 2023

Found helpful?