From: ABYOLOv4: improved YOLOv4 human object detection based on enhanced multi-scale feature fusion
Convolution kernel
Dilation rate
Receptive field
3 × 3
1
3
2
5
7
4
9
11