Skip to main content

Table 1 Layered fusion implementation details

From: ABOS: an attention-based one-stage framework for person search

Attention network layer n×c×h×w,attention
\({P_1}\) 1×768×h×w,A,M,3×3C,Sigmoid
\({P_2}\) 1×1536×h×w
\({P_3}\) 1×3072×h×w,A,M,1×1F,ReLu,1×1F,Sigmoid
  1. n denotes the number of filters, c denotes the channels of the filter, h/w denotes the size of the filter, A denotes the average pooling layer, M denotes the maximum pooling layer, C denotes the convolutional layer, and F denotes the fully connected layer