CUSTOM AI
Last updated
Last updated
CUSTOM AI is a function that conducts manual labeling of objects you want to label and then auto-labels with an artificial intelligence model created based on this.
At least 10 manual labeling per class is required to start auto-labeling.
On the Dashboard tab of the labeling project, click the Create CUSTOM AI
button at the bottom right of the Tab menu.
(If the Create CUSTOM AI
button is not enabled, check if any of the classes have less than 10 labelling cases.)
Under Select CUSTOM AI type, set the labeling type for CUSTOM AI.
Under Select the class to use, select all classes that will be utilized for CUSTOM AI generation and auto-labeling.
Click the Start AI Training button at the bottom to training learning CUSTOM AI.
You can check the start and completion of the learning by email and the notification tab.
When CUSTOM AI learning is complete, click the Select Best Custom AI
button.
Check the object recognition output of each generated model and select the optimal result (redundant selection is possible)
Click the icon in the upper right corner of the object Detection output image to view the image zoom.()
Finally, click the Complete
button to complete the model selection.
On the Dashboard tab of the Labeling project, click the Start Auto Labeling
button at the bottom right of the Tab menu.
Select CUSTOM AI under Select AI type, and select the model you want to perform auto-labeling on.
Select the number of images to perform auto-labeling and click the Start Auto-labeling
button at the bottom.
You can check the progress of auto-labeling on the Dashboard tab, and you will be notified by e-mail and notification when it is completed.
In the Labeling Project's Data List tab, set the filter at the bottom right of the tab menu to Complete Task Status.
Click the data you want to view in the Data List to see the auto-labeling results, and click the Inspect Label
button in the bottom right to modify the labeling.
(For more information, see Labeling > Manual-Labeling > Object Detection.)
With Inspectation and Modification, CUSTOM AI models can be regenerated with high-quality data, allowing more data to be auto-labeled more accurately. Repeat this process to continue improving auto-labeling.