基于动态采样对偶可变形网络的实时视频实例分割

宋一然,周千寓,邵志文,易冉,马利庄

Dynamic sampling dual deformable network for online video instance segmentation

Yiran SONG,Qianyu ZHOU,Zhiwen SHAO,Ran YI,Lizhuang MA

表 3 基于YouTube-VIS 2021验证集的精度比较

Tab.3 Accuracy comparison based on YouTube-VIS 2021 validation set

方法 mAP/% AP50/% AP75/% MaskTrack-RCNN [9] 28.6 48.9 29.6 SipMask [2] 31.7 52.5 34.0 CrossVIS [3] 34.2 54.4 37.9 IFC [22] 36.6 57.9 39.3 DSDDN 34.8 55.9 37.4