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 8 关于Transformer块层数的消融学习

Tab.8 Ablation study on number of layers in transformer blocks

层数 mAPDTE/% mAPCATE/% mAPDTD/% 1 34.7 36.5 35.2 2 36.1 36.9 37.7 3 36.5 37.3 37.1 4 36.6 37.7 37.3 5 36.3 35.9 37.5 6 36.6 35.4 37.4