Earf-YOLO: An Efficient Attention Receptive Field Model for Recognizing Symbols of Zhuang Minority Patterns

Figure 5

(a) is the SRFB during training and (b) the SRFB during prediction. According to the additivity of convolution, the SRFB can treat 1 × 1, 3  × 1, and 3 × 1 convolution (consisting of a large number of zeros in the convolution kernel) as a special 3 × 3 convolution.

(a)

(b)