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

<div>Above is RFP. Below is SLFF. SLFF uses the output of the backbone network as the input of FPN to obtain the multiscale feature map. The self-recurrent module incorporates feedback connections into FPN. RFP connects them back to the bottom-up backbone. SLFF lets them connecting back to the FPN. Predict module uses the prediction part of YOLOv3.</div>

Mathematical Problems in Engineering

fig2

Figure 2

Figure 2: Self-Recurrent Learning and Gap Sample Feature Synthesis-Based Object Detection Method 