BRNet: Bi-residual network with aggregate connection

Yongwei Liao, Zhenguo Yang, Wenyin Liu

Abstract: Highlights•Propose a bi-residual module by extending residual block with aggregate connection.•Devise an aggregate connection consisting of average pooling, Relu, 1 × 1 convolution and the Softmax components.•Verify the superiority of the BRNet on image classification and object detection.