Ground truth refers to the accurate, real-world values or labels of data, used as a gold standard for training or evaluating models. For example, in an image dataset for object detection, the ground truth might consist of the labeled bounding boxes and classes of objects in each image (as annotated by human labelers). Models are trained to approximate the ground truth mapping from inputs to outputs, and evaluation metrics are computed by comparing the model’s predictions against the ground truth. The term is used across machine learning and remote sensing (“ground truth” might literally mean measurements taken on the ground to validate aerial or satellite data). The quality of ground truth directly affects model performance – noisy or incorrect ground truth can lead the model astray. Establishing reliable ground truth can be challenging and often involves significant human effort or high-precision instruments.
Data Selection & Data Viewer
Get data insights and find the perfect selection strategy
Learn MoreSelf-Supervised Pretraining
Leverage self-supervised learning to pretrain models
Learn MoreSmart Data Capturing on Device
Find only the most valuable data directly on devide
Learn More