Training Data Cleaning Project
SuperMap iDesktopX provides a functional module for efficiently managing, editing, and preparing sample data for AI model training. It aims to help users systematically organize, classify, annotate, and correct large-scale image training datasets to improve data quality and lay a solid foundation for subsequent model training.
Samples
In machine learning tasks, a sample refers to an independent data unit and its corresponding reference truth value. In remote sensing imagery interpretation, a sample usually refers to a single image or image patch paired with label data.
All Positive Samples
Refers to samples where all pixels in the image belong to the target type. Such samples do not contain background or other categories.
All Negative Samples
Refers to samples where all pixels in the image do not belong to the target type, meaning all pixels belong to the background or other categories.
Label
Reference truth data corresponding to the image pixel by pixel or object by object. It defines the semantic category to which the sample belongs, the location boundaries of targets, or other attributes, and is the target output for model learning.
Positive Label
A label marked as the target type.
Negative Label
A label marked as background or non-target type.
Related Topics
Samples Category and Label Editing