1 paper
Xiaogang Du, Dongxin Gu, Tao Lei +2
In recent years, encoder-decoder networks have focused on expanding receptive fields and incorporating multi-scale context to capture global features for objects of varying sizes.…