We thoroughly reviewed the building object recognition literature with the intent of quantifying the state-of-the-art recognition performance for each building object class. In total, 63 papers were included in the summary plot (below).
Width of the bars is proportional to the number of object instances in the validation dataset of the study. The vertical axis summarizes each method’s performance with an F-score.
In order to have your object recognition method included, email the authors with the study citation and associated Uniformat classes, number of instances, and F-score.