The term ground truth originates primarily from the fields of artificial intelligence, big data and automation. Ground truth literally means "true foundation". It refers to a type of reference or comparative data that is considered to be absolutely correct. This data is used to check the results of algorithms or models.
Imagine you are developing an image recognition system for self-driving cars. To test how well the system recognises road signs, you need a set of images in which a human has already entered exactly where which road sign is located. These images marked as correct are the ground truth. Your system now compares its results with this reference data. The better the match, the more reliably your system works.
Ground truth is therefore crucial for training artificial intelligence or evaluating the accuracy of automated processes. Without such reliably known comparative data, it would hardly be possible to measure and further develop progress in areas such as image recognition or machine learning.