Semi-automatic labeling reduces the human effort in generating a training data set for classification and segmentation tasks. If you want to classify data points - pictures for example - into categories, you must manually classify some of them first. This activity is known as labeling. Labeling each image costs you the same amount of effort, but the information provided by each label is not the same. Semi-automatic labeling helps you label the images that add the most information and skip those that add little information. In total, the amount of labor is reduced by over 80%.
Check out the video below for a full explanation!