CSPNeXt: A new efficient token hybrid backbone

Xiangqi Chen, Chengzhuan Yang, Jiashuaizi Mo, Yaxin Sun, Hicham Karmouni, Yunliang Jiang, Zhonglong Zheng

Abstract: Highlights•The high and low-frequency information are combined in an efficient manner.•The local and global feature representations are balanced in a new manner.•A new channel segmentation mechanism is introduced.•The proposed architecture achieves good performance on several vision tasks.