Road Sign Recognition Method Based on Segmentation and Attention Mechanism

Table 1

ERFNet model architecture distribution.

	Serial number	Module type	Number of output channels	Down-sampling multiple
Encoder	1	Downsample block	16	2
	2	Downsample block	64	4
	3–7	Non-bottleneck-1D × 5	64	4
	8	Downsample block	128	8
	9	Nonbottleneck-1D(2-dilated)	128	8
	10	Nonbottleneck-1D(4-dilated)	128	8
	11	Nonbottleneck-1D(8-dilated)	128	8
	12	Nonbottleneck-1D(16-dilated)	128	8
	13	Nonbottleneck-1D(2-dilated)	128	8t
	14	Nonbottleneck-1D(4-dilated)	128	8
	15	Nonbottleneck-1D(8-dilated)	128	8
	16	Nonbottleneck-1D(16-dilated)	128	8
Decoder	17	Upsample block	64	4
	18–19	Nonbottleneck-1D × 2	64	4
	20	Upsample block	32	2
	21–22	Nonbottleneck-1D × 2	32	2
	23	Upsample block	C	1