Self-Recurrent Learning and Gap Sample Feature Synthesis-Based Object Detection Method

<div>REAML. This figure unrolled the self-recurrent structure to a 2-step sequential network. Dilated convolution is used to transform the size of the feature image and fuse it. Note: ©, channel splicing operation; ⊕, feature fusion for automatic weight allocation.</div>

Mathematical Problems in Engineering

fig4

Figure 4

Figure 4: Self-Recurrent Learning and Gap Sample Feature Synthesis-Based Object Detection Method 