Multiscale Efficient Channel Attention for Fusion Lane Line Segmentation

Figure 4

The structure of efficient channel attention. The input channel here is 64.