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

Table 5

Comparison of detection results based on the original training dataset and downsampled dataset.

Class	Object number in the training dataset	mAP (original)	mAP (downsampled)
Hatchback	12165	0.669	0.744
Sedan	5484	0.573	0.565
Minibus	3220	0.663	0.601
SUV	1761	0.560	0.576
MPV	898	0.154	0.142
Pickup	263	0.020	0.0001
On average	3965.2	0.440	0.438