Image Annotation Tool – Clickworker Instructions

Goal: The image annotation tool is used in projects where the aim is e.g. to mark silhouettes of people, tag products in shelves or segment radar images. Most data done with the image annotation tool is used to train AI systems.

This guide explains in a few short steps existing features and how to use the annotation tool.

1. Feature: Zoom

Sometimes in order to get a closer look at details on a picture and do your work more accurately, a zoom is required. You can zoom in and out of a picture from 10 to 500 %.

Be careful: When you zoom out again, sometimes tags relocate. Zoom further out to get them in the right position again.

2. Feature: „Not doable or duplicate photo“

If there is an image series in your job (not just one picture to be tagged, but several) you can mark single images with a red cross by using the “not doable” feature. It helps to blend out duplicate images that might otherwise confuse you or to simply disregard pictures where no tagging is necessary given the current instructions.

3. Feature: „Show all tag hovers“

This feature allows you to see already marked categories/tags to get a better overview of what might be still missing.

4. Feature: „Monochrome“

This feature allows you to see images in bluish colors. It is especially helpful when there are white colors in the image, that normally might prevent you from seeing the white mark points you made in the picture.

5. Feature: “Reset”

Please use this feature with caution. This button can be used to reset all the annotations you have done in a job. Therefore, please don’t use it to reset single elements!

6. Feature: “Superpixels”

This tool locates color differences on its own and automatically marks parts of the image as a polygon. Unfortunately, the AI is not working precisely enough yet, so there is no need to explain this feature in detail at this point.

Please don’t use this feature unless it is explicitly required in the briefing.

7. Use of “Keyboard shortcuts”

Please read this information carefully, when you are using the tool for the first time.

A) Removing elements: To be able to remove elements your cursor has to hover above the element you want to remove while you use the key „Delete“ or „Backspace“.

B) Editing elements: The edit mode (key: „e“) will help you to edit and adjust elements. Use: mouse cursor over element + key: “e”

C) Relocating elements: The selection mode (key: „s“) will help you to relocate elements when you have set them incorrectly. Use: mouse cursor over element + key: “s”. Attention: If you accidentally move part of an element out of the image, press key: ESC to delete the element completely and start from scratch.

8. Feature: “Rectangles”

The rectangle is used for the annotation of elements that are either always rectangular or where only width and height and not the exact margins are relevant. It is a less exact type of annotation compared to e.g. polygons. Please observe the detailed briefing so you’ll know when to use rectangles.

9. Feature: “Polygons”

A polygon is used when the object that requires annotation, should be tagged very carefully for polygons deliver the most exact results. It is often used when the exact shape of the to be tagged elementsis relevant. Example: to mark a person’s silhouette in detail.

How to use polygons: Make sure to start at one point and trace the object that should be tagged carefully by setting dots around the borders. These dots connect automatically to fine lines and create a full shape once you link the last dot with the very first dot you have set.

How to tag several objects in one picture: Sometimes you will be asked to tag more than one object/category on a picture. It can happen that some polygons overlap each other. You will not be able to make changes or adjustments to the first tag you did unless you hide the last one first, which ‚lies on top’ of the first polygon. (also please see point 11 – Eye Sign).

Once you complete tagging the first element and want to proceed with tagging the next object, you cannot choose a starting point which lies within the first tagged element as this would change the category of it. However, when you hide the first tagged element (see again point 11 – Eye Sign) you can start tagging your next object.

10. Features: “Crosshairs” / “Tags”

Crosshairs and tags are used for very simple annotations. Example: You are looking at a picture where you see many objects of the same kind. You place a tag (dot) on each object and can later on see how many objects are on display in this picture. This type of tagging is very helpful for object counting or to simply classify two or more category types.

11. Eye sign

This eye sign is a helpful tool in order to get a better overview of your work or to access specific tagging elements. You simply hide elements or categories by clicking on this tool. It is particularly helpful when elements overlap and it becomes hard to see the boundaries of single elements.

12. “Categories”

Categories are labels for the objects you see. Each element that is tagged needs a category and some perhaps even subcategories. When there are subcategories it is absolutely necessary to select both the main category and subcategory!

Example:

alphabetical letter > consonant

alphabetical letter > vowel

How to tag categories:

Possibility 1:You can select the tag type, category and subcategory first before annotating the respective object.

Possibility 2:You can select the tag type and the category first, then tag the element first before you decide on a subcategory.

Selected a wrong category? No problem, you can always adjust them again afterwards for already completed annotations. Simply click on the element in question and review your options.