Figure 5 | Scientific Reports

Figure 5

From: Deep leaning-based ultra-fast stair detection

Figure 5

Illustration of the SE-ResNeXt block with dilated convolution. The original standard group convolution with 32 groups is divided into two branches with 16 groups, and the dilation rates of the dilated convolutions are (4,16) and (16,16). Finally, the results of the two branches are concatenated. A layer is denoted as (input channels, filter size, output channels).

Back to article page