To understand the effect of object detection efficiency in terms of dataset which has the large interclass object size differentiation among classes, this work validated 11 classes of dataset with larger interclass object-size disparity which were categorized into three groups including the mission-target group, the protective group, and the attentive group for training and validation by three popular detectors, that is, YOLOv3, YOLOv4, and YOLOv5. The experimental result reports that the class of the smaller-size object size although involved the largest number of labeled ground-true objects in training set among all classes, yet the Average Precision (AP) performed the worst one, on the contrary, the bigger object size although engaged the fewer number of training set, in which AP are still given the promising performance, of which the influential factors and analysis are discussed in the experimental result.