Efficient BFCN for Automatic Retinal Vessel Segmentation

Figure 5

The changing process of the receptive field, (a) the changing process of the ordinary convolution receptive field with a two-layer convolution kernel of 3 × 3, stride of 1, and padding of 0, and (b) the changing process of the dilated convolution receptive field with the convolution kernel of 3 × 3, stride of 1, padding of 0, and expansion rate of 2.

(a)

(b)