Fig. 8From: Deep learning-based automatic-bone-destruction-evaluation system using contextual information from other jointsThe difficult case of erosion for all four classification models. Each bounding box represents the detected target joints. The color of the bounding box indicates how the classification results compare to ground-truth (GT). Blue indicates GT is non-intact, and prediction is non-intact (true positive). White indicates GT is intact, and prediction is intact (true negative). Yellow indicates GT is intact, and prediction is non-intact (false positive). Red indicates GT is non-intact, and prediction is intact (false negative). GT, ground-truth; SISO, single-input single-output; MIMO, multiple-input multiple-outputBack to article page