Fast and Accurate Hand Visual Detection by Using a Spatial-Channel Attention SSD for Hand-Based Space Robot Teleoperation

Figure 4

Multiattention module with feature fusion. The structure is the MA-FF on conv4_3 layer. In practice, the output of Deconv-1 adds with conv7, and the output of Deconv-2 adds with conv4_3. + means the symbol of element wise add operation, and × means the corresponding element multiply operation.