Fig. 7From: Anchor-free object detection with mask attentionThe performance of the proposed method is on the three scales of large, medium and smallBack to article page