Annotation Tools#
In this section, we will walk you through the tools you will be using for annotating images and video frames. The use of the tools is very intuitive, however, there are some functionalities that are not obvious straightaway. We will explain what these tools do, how to use them, and when to use them.
Below, you will find a list of all the tools you will have at your disposal. However, you cannot use all of them for each type of project. They are computer-vision-task-bound. What it means is that not every tool is available for all projects e.g. a detection project requires a tool different than that in a classification project.
Basic annotation tools |
|
|---|---|
|
A bounding box tool is most aptly associated with an object detection task. The bounding box is a rectangle or a square drawn surrounding an object in an image. Click here to learn more about the bounding box tool. |
|
A rotated bounding box tool is an extension feature of the bounding box that encloses an object within a minimal rectangle. Click here to learn more about the rotated bounding box tool. |
|
A circle tool is available for an image segmentation task. Its purpose is to simplify the annotation of circular objects. Click here to learn more about the circle tool. |
|
A polygon tool is solely dedicated for image segmentation tasks. The polygon tool allows for free form drawing around shapes. Click here to learn more about the polygon tool. |
Smart annotation tools |
|
|---|---|
|
A quick selection tool is intended to simplify image segmentation. Simply draw a bounding box around the object and Intel® Geti™ will fit a polygon to the shape of the object. Click here to learn more about the quick selection tool. |
|
An object coloring tool is also intended to simplify image segmentation. Simply select the brush and draw over the objects you want to annotate. Click here to learn more about the object coloring tool. |
|
A detection assistant tool is also intended to simplify image segmentation. Simply select the SSIM icon and draw a bounding box or a circle on an object and the system will highlight similar-looking objects to the one you annotated. Click here to learn more about the detection assistant tool. |
|
An interactive segmentation tool is also intended to simplify image segmentation. Simply select the interactive segmentation icon and embed a dot mark inside an object, the annotation tool will automatically annotate the object of interest. Click here to learn more about the interactive segmentation tool. |
|
An automatic segmentation tool simplifies the process of segmenting objects within images. By hovering the mouse cursor over an object and embedding a dot mark within it, you can preview and accept an annotation. Click here to learn more about the automatic segmentation tool. |
Basic annotation tools#
A set of basic annotation tools is geared towards you having full control of how you want to annotate objects in an image or video frame. However, it requires more manual labor as opposed to using our smart annotation tools. What is more, these tools are also tied to specific projects e.g. from the set of basic annotation tools, you can use a bounding box tool in an object detection project whereas in an image segmentation project you can use all of the tools. The reason why some tools are available for specific projects pertains to the technical aspects of computer vision tasks. You can read more about it in our AI fundamentals.
Selector tool#
The selector tool is available for detection and segmentation tasks. You can select a bounding box or a polygonal shape and edit it - change the size and shape of the annotation and/or change a label for that annotation. You can also select a region with annotations and automatically change a label for all of them, hide, lock, or remove them all.
The selector tool has two editable modes for changing the size and shape of the annotation:
edit by point - where you can drag and move vertices of the annotation and move them by dragging
edit by brush - where you push vertices inwards and outwards
The way edit by brush mode works depends on where you position your cursor. If you click on a region outside the annotation, you will be able to subtract the annotated region. If you click on a region within the annotation, you will be able to add a region to the annotation. You can resize the brush by moving the slider to the left or right.
You can create and reuse annotation templates from existing annotations, called stamps. To create a stamp, you can choose one of the ways below:
right click on the annotation and select Create stamp
select the annotation and go to the Selector tool menu to choose the stamp feature
from there
To cancel the stamp mode, you can choose one of the ways below:
press “Esc” on your keyboard
click on
in the Selector tool menuselect a different tool
Bounding box tool#
A bounding box is most aptly associated with an object detection task. However, an image segmentation project can also utilize a bounding box tool. The tool is activated by default in the detection project.
To activate the tool in a selection project, click on 
from the left-hand toolbar.
The bounding box is a rectangle or a square drawn surrounding an object in an image. The bounding boxes are used for identifying objects in the image and localizing these objects. These boxes give a computer the coordinates through which it can find the objects in the image or video. This form of annotation is best used for objects of uniform, regular shape e.g. a road sign. What this annotation tool is not good for is annotating objects that overlap e.g. cells, atoms, etc.
Once you draw the bounding box:
You can resize the box immediately after drawing it.
Select a box with a selector tool
to move or resize the box.Select a box with a left click of the mouse button to move or resize the box.
After drawing the box, you can label it straightaway or keep drawing the boxes and leave the task of selecting the labels for last. If you want to remove multiple boxes at once, left click and drag the area with the boxes you want to remove. Then, click on 
.
Rotated Bounding Box#
Rotated Bounding Box is a rectangle or a square drawn surrounding an object in an image. Compared to a standard bounding box tool, you can specify the angle/rotation for the bounding box and enclose it so that it does not take up unnecessary space. You can rotate the box by clicking on this icon 
above the bounding box.
To better illustrate the difference between a regular bounding box and an oriented bounding box, look at the picture below.
Bounding Box |
Rotated Bounding Box |
|---|---|
|
|
When an object is long and is positioned at an angle (e.g. 45 degrees) then a standard upright bounding box takes up a lot of empty space with the object occupying just the diagonal. A rotated box is much more efficient and cleaner way of highlighting the object of interest.
Take note of how the standard bounding boxes overlap unnecessarily in the image above.
Hint
As rotating the bounding box may require a lot of manual precision and time, you can leverage the Interactive Segmentation tool inside Oriented Detection projects to speed up the annotation process.
Circle tool#
A circle tool is available for an image segmentation task. So you will have access to this tool, if you create a segmentation project.
To activate the tool, click on the circle tool icon 
from the left-hand toolbar.
To annotate a circular object, just hover on the object and click the left button of your mouse. If you need a bigger circle to annotate the object, you can hold the left mouse button and move your mouse to resize the circle or click on the white circle on the edge of the circle annotation to resize it.
Although, there is another, recommended, way to resize the circle radius. You can right click the mouse button, hold it, and move your mouse to resize the radius. The circle then will remain the same size for further annotation. You can also resize the radius of the circle by moving the slider in the top tool bar.
If you want to remove multiple annotations at once, select the Selector tool , and left click and drag the area with the boxes you want to remove. Then, click on the remove icon .
Hint
Resize the radius of the circle to the smallest or very small radius. Thanks to this operation, you will be able to annotate the circular objects more precisely.
Polygon tool#
A polygon tool is solely dedicated for image segmentation tasks. You can only access the tool in a segmentation project. The tool is activated by default in the segmentation project.
The polygon tool allows for free form drawing around shapes. The tool is ideal for annotating irregular shapes and provides shape delineation of pixel-size precision. You can use this tool for the projects that require complex annotations e.g. annotating medical imaging.
To use this tool, click on this icon |polygon icon| and click on an image to draw your polygon. A blue circle will be drawn indicating where the start point is. Continue clicking around the shape or hold the left mouse button to draw around an object in one continuous motion. To finish your polygon, click or hover on the start point (the blue circle) of the polygon.
To help you automatically detect edges, toggle Snapping mode in the top toolbar. The mode draws delineates an object by drawing lines around it. While hovering around the edge of the object, remember to click on the edge to save the state of the object delineation. The mode may not always delineate the entire shape in one continuous movement.
If you want to change the shape of your polygon right after you draw it, click on vertices to manipulate the shape. To edit previously drawn polygons, choose the selection tool and click on a polygon annotation or left click and drag an area with the polygon annotation.
You can also extend the shape by manipulating vertices. To do so, hold and drag the vertices of the annotation to a desired polygon shape.
Smart annotation tools#
A set of smart annotation tools is intended to automate your annotation labor. Creating annotations can be tedious and time-consuming if you have thousands of objects to annotate. The smart annotation tools reduce the annotation time and effort multi-fold.
Furthermore, these tools are solely dedicated to an image segmentation project. The reason why some tools are available for image segmentation, pertains to the technical aspects of computer vision tasks. You can read more about it in our AI fundamentals.
Object selection tool#
An object selection tool belongs to our smart annotation tools, which means your annotation efforts are automated by Intel® Geti™. This tool is available for an image segmentation task, accessible in a segmentation project.
To activate the tool, click on the quick selection tool icon 
from the left-hand toolbar.
After selecting the tool, draw a bounding box around the object and the system will fit a polygon to the shape of that object. Sometimes, Intel® Geti™ may not create a perfect delineation around the object - that is why you will find the foreground icon 
and the background icon 
for the object selection tool in the top toolbar.
The background tool allows you to erase the parts of the shape that belong to the background. For illustrative purposes, a scissor analogy will be used to help you understand how this tool works. Imagine drawing a circle on a piece of paper and cutting this circle out of the piece of paper you drew it on. The background tool is your pair of scissors used for cutting out the drawn circle.
Similarly, by selecting the foreground tool , you can draw to add parts of the object to the annotated shape.
You can also influence the precision with which the Intel® Geti™ platform automatically annotates your selected object. To adjust precision, choose a number from the Sensitivity drop-down menu in the top toolbar. Higher numbers increase precision, while lower numbers decrease it. Be aware that higher precision requires more computing power and time. Adding high-resolution images may significantly prolong the annotation waiting time.
Object coloring tool#
An object coloring tool belongs to our smart annotation tools, which means your annotation efforts are automated by the Intel® Geti™ platform. This tool is available for an image segmentation task, accessible in a segmentation project.
To activate the tool, click on the object coloring tool icon 
from the left-hand toolbar.
After selecting the tool, choose a label - “background” and draw line(s) on the area that constitutes the background of objects you want to annotate. After this, draw lines on the objects you want to annotate. The polygon will be automatically drawn around the object. If the tool does not annotate the whole object, keep drawing the lines until you are satisfied with the result.
Alternatively, you can change the order of selecting labels and start with drawing lines on the objects first and then marking the background.
You can also influence the precision with which the Intel® Geti™ platform automatically annotates your selected object. You can also influence the precision with which the Intel® Geti™ platform automatically annotates your selected object. To adjust precision, choose a number from the Sensitivity drop-down menu in the top toolbar. Higher numbers increase precision, while lower numbers decrease it. Be aware that higher precision requires more computing power and time. Adding high-resolution images may significantly prolong the annotation waiting time.
Detection assistant tool#
A Structural Similarity (SSIM) annotation tool is a smart annotation tool that highlights similar-looking objects to the one you annotated by drawing a bounding box or a circle on an object.
To activate the tool, click on the SSIM tool icon 
from the left-hand toolbar.
After selecting the tool, you can choose a shape (rectangle or circle) which you want to draw from the top toolbar. Once you annotate an object, the Intel® Geti™ platform automatically detects and annotates objects that look similar to the one you annotated.
The merge duplicates feature is turned on by default, which prevents the Intel® Geti™ platform from creating the same annotations for the similar-looking objects.
You can also influence the precision with which the Intel® Geti™ platform automatically annotates similar-looking objects. You can also influence the precision with which the Intel® Geti™ platform automatically annotates your selected object. To adjust precision, choose a number from the Sensitivity drop-down menu in the top toolbar. Higher numbers increase precision, while lower numbers decrease it. Be aware that higher precision requires more computing power and time. Adding high-resolution images may significantly prolong the annotation waiting time.
Interactive Segmentation tool#
The Interactive Segmentation tool is a smart annotation tool that automatically annotates the object of interest by embedding a dot mark inside the object.
You can trigger positive/negative points with the left and right mouse click respectively. Positive points (left mouse click) add the annotation region to the object of interest while negative points (right mouse click) subtract the annotation region.
You can also leverage the Dynamic selection mode which allows you to limit the area in the media item for the algorithm to find the object of interest within the defined boundary of that area. You first need to draw a bounding box around a particular area and then left click on the object of interest.
Automatic Segmentation tool#
An automatic segmentation tool is a smart annotation tool that automatically annotates the object of interest by hovering a mouse cursor over the object to preview the annotation and embedding a dot mark inside the object to accept the annotation.
Once you create a project, the automatic segmentation tool will be available in the tasks bar on the left pane of the screen. When you click on 
the Intel® Geti™ platform will start extracting the image’s features.
Note
Extracting information from the image may take a few seconds to make annotations preview appear instantly when you hover your mouse cursor over objects.
Hover your mouse cursor over an object in the image to preview the polygon that can be generated. The preview is shown with dashed lines while clicking on the preview solidifies the selection.
Modify annotations using positive (green) and negative (red) points for complex objects. With the Interactive mode:
When turned off: Placing a point on the canvas immediately accepts the shape, adding the annotation. No need to press accept or add another point.
When turned on: Place a point to see a tool-generated shape preview. If satisfied, click accept. You have the option to add positive or negative points (with the Righ-click mode on or by clicking inside of the shape with Right-click mode turned off). After each click, the shape is recomputed, giving choices to accept, reject, or refine.
To help visualize which parts of the image are being loaded or not, adjust the Mask capacity setting. A setting of 0% will show the image in its original color. Higher settings will emphasize the areas of the image that are being annotated.
The tool changes its output based on the task you annotate:
Oriented Detection: Produces a rotated bounding box for objects that are at an angle, providing a more accurate representation.
Detection: Creates a standard bounding box that encloses the object, used for more straightforward detection tasks.
Segmentation: Generates polygon annotations, allowing for precise outlining of complex shapes and objects.
Hint
It’s recommended to disable labels when using this tool if the image contains many small objects. Disabling labels can provide a cleaner view, making it easier to work with intricate details and numerous elements.