The term „ground truth“ originates primarily from the fields of artificial intelligence, big data, and automation. Ground truth literally translates to "true basis." It refers to a type of reference or comparison data that is considered absolutely correct. This data is used to verify 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 traffic signs, you need a set of images where a human has already precisely marked the location of each traffic sign. These correctly marked images 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 for evaluating the accuracy of automated processes. Without such reliably known comparison data, it would be hardly possible to measure and further develop progress in areas such as image recognition or machine learning.
Deutsch 
English (UK)