ANNOTATION_GUIDE.md

LabelBox Annotation Guide:

1. Create new project:

1. Frist Go to and create an account.
2. After successful sign-up/sign-in, under the project tab, click on new project to create a new project.
3. Give a name to the project.
4. Under Attach Dataset section, choose the upload new dataset to upload the dataset.
5. Then give a name to the dataset.
6. Under Customize your labeling interface section, choose the SELECT EXISTING and select Image Labeling.

2. Configure Interface:

2. Configure Interface:

1. Delete all the existing object by clicking on the delete icon of the object on the side panel.
2. Now click on Add Object , give name as line-border and select line as annotation method.
3. Again click on Add Object, give name as side-text and select Polygon as annotation method.
4. Delete all the classification under the Classifications section on the side panel.
5. Now click on confrim.
6. Then click on the COMPLELTE SETUP.
7. Finally click on Start Labeling to start the Labeling/Annotation process.

3. Annotaion/Labeling Procedure:

line-border:

1. Select line-border annotation method and start labeling as polylines separating various text lines.
2. Annotation guide:
   • Zoom in the image for better labeling.
   • Start from left to right.
   • The starting and ending point of polyline (line-border) should not touch the border of Pecha.
   • Select a point just near the text without touching them [Tips: Zoom-in for better visual] except for the place where top and bottom characters are touched, select an appropriate point can separate out the two characters.

Side-text:

1. Select line-border annotation method and start labeling as polygon to surround the side text.
2. Annotation guide:
   • Zoom in the image for better labeling.
   • Construct a polygon that can perfectly surround the side text.

After the labeling is done for one image, click on SUBMIT at bottom of side panel to save the label and proceed to next labeling.

4. Export the label:

1. Click on the home button at top of the side panel.
2. Click the Export tab, select Export Type as JSON and Label Format as XY and set Generate Mask ON.
3. Finally click on GENERATE EXPORT.
4. To download the label file, Click on the Tasks as top-right and click on the download icon.

Simple Example of labelbox annotation:

More Details example of annotations:

• Please try to annotate the dataset as the following examples.
• Download the following examples images for better look at places where the top and bottom characters are touched.