Conceptual Cognitive Modeling for Fine-Grained Annotation Quality Assessment of Object Detection Datasets

Table 7

Annotation accuracy of ground truth annotations for the selected images of the PASCAL 2007 dataset (the average values are computed for the twenty classes).

Class	Flat accuracy of label		Hierarchical accuracy of label
	Precision	Recall	Precision	Recall
Person	0.97 ± 0.01	0.79 ± 0.11	0.98 ± 0.01	0.81 ± 0.1
Car	0.94 ± 0.02	0.82 ± 0.08	0.96 ± 0.01	0.83 ± 0.08
Chair	0.90 ± 0.02	0.81 ± 0.09	0.96 ± 0.01	0.84 ± 0.08
Bottle	1 ± 0	0.74 ± 0.10	1 ± 0	0.81 ± 0.09
Potted plant	0.98 ± 0.02	0.77 ± 0.02	0.99 ± 0.01	0.78 ± 0.03
Cow	0.99 ± 0.01	0.81 ± 0.10	1 ± 0	0.83 ± 0.10
Dining table	0.68 ± 0.16	0.63 ± 0.11	0.87 ± 0.06	0.65 ± 0.09
Bus	0.91 ± 0.09	0.89 ± 0.08	0.94 ± 0.05	0.90 ± 0.07
On average	0.94 ± 0.02	0.84 ± 0.05	0.97 ± 0.01	0.87 ± 0.05