Labeling your images
Make sure to follow the best practices recommendations for labeling listed in the subpage Labeling best practices
From the Dashboard, click any image to enter into the Labeler. The Dashboard updates each time you annotate an image. Finished images will appear in the right column (Labeled), and the images you are currently labeling will appear in the middle column (In progress).
Once you click on an image, you will enter into the Labeler.
From left to right, the labeler has three sections:
- 1.Image column, that allows you to switch quickly between images and see their status.
- 2.Labels column, that allows you to add object and region classes, and see currently labeled objects in the image.
- 3.Zoomable label view, which allows you to draw labels easily, zooming in and out of the image you are labeling.
The top bar allows you to mark the image as completed, as well as adjust brightness and contrast for the image to help you label more effectively.
A class is simply a biological (or non-biological) object that you see in the object. You could create classes to denote regions of inflammation, individual cells, structural regions, organoids, and more.
Once you are on the labeling screen, simply click on the "+" sign next to Object classes to create a new object class.
After naming, select the segmentation type for your class: Detect each object (instance segmentation) or Detect each region (semantic segmentation). You can read about the differences between instance and semantic segmentation in our blog post about cell segmentation.
Not sure whether to pick object or region segmentation? In most cases, choosing Object Detection is the right choice and gives richer metrics on a per-object basis.
After creating the necessary classes, you can start labeling the images. To start annotations, use the keyboard shortcut on the right side of the class you created to start drawing. Alternatively, click on the class you want to annotate. You'll now enter into drawing mode. At any time, hit
ESC to come out of object drawing mode.
Using the pen tool (press p), you can click and drag in order to create a labeled object, as shown below.
Switch to Autolabel by pressing b on your keyboard. Then, drag a box around the object you are looking to detect. For best results, try to keep your box fairly close to the object but not too tight around the object.
Once you let go of the box, you will see a Detecting... pop up and an object will appear. The model has been trained on thousands of biological images to identify what it thinks is the object and draw a polygon around it.
Once an object is detected, you have around 5 seconds to click the convenient X button on the top left to delete the object without leaving the Autosegment tool.
- Changing the brightness and contrast of the image using the sliders may improve the performance of Autolabel
- Playing with the size of the box you draw around the object can change the performance of Autolabel
- If you cannot get Autolabel to work for your images, send us a message through the bottom right and we'll see if we can integrate your images into training.
Simply hold
Shift to hide all objects temporarily, to help you label effectively.In the Labeled Objects section, click on the open eye icon to hide objects. If you click underneath a grouped class, it will hide every object under that group. If you click on the eye icon next to a single object, you can hide just that object from view.
Make sure that you are out of drawing mode (press
ESC). Select an object by clicking it in the viewer or from the Labeled objects section. Then, press the key e or click on the pencil in the Labeled objects section to enter editing mode. From here, you can choose to Add or Subtract from the object. Draw an object that will be added or subtracted from that object. Once your object has been completed, your changes will be automatically saved.Once you are finished with an image, you can mark it as completed by using the shortcut
Shift+Enter or by clicking the Mark as completed button on the top bar.From the top bar, click the Channel selector. Here, you can adjust brightness and contrast on a per-channel basis, or hide a specific channel for an image.
Depending on the segmentation type, your created classes appear under Object classes (for instance segmentation) or Region classes (for semantic segmentation).
You can also choose from different colors to represent the class. Simply click the colored circle to open up a color picker.
Hover over the class to reveal the 3 dots button, which you can click. From there, use the dropdown to rename or delete a class. If you delete a class, every object created for that class will be deleted as well.
Last modified 3mo ago