Two-Stage Intelligent DarkNet-SqueezeNet Architecture-Based Framework for Multiclass Rice Grain Variety Identification

Table 1

Description of SqueezeNet architecture.

Layer name/type	Output size	Filter size/stride(if not a fire layer)	Depth	S1 × 1 (#1 × 1 squeeze)	e1 × 1 (#1 × 1 expand)	e3 × 3 (#3 × 3expand)
Input image	224 × 224
Conv1	111 × 111 × 96	7 × 7/ (x96)	1
Maxpool1	55 × 55 × 96	3 × 3/2	0
Fire 2	55 × 55 × 128		2	16	64	64
Fire 3	55 × 55 × 128		2	16	64	64
Fire 4	55 × 55 × 256		2	32	128	128
Maxpool4	27 × 27 × 256	3 × 3/2	0
Fire 5	27 × 27 × 256		2	32	128	128
Fire 6	27 × 27 × 384		2	48	192	192
Fire 7	27 × 27 × 384		2	48	192	192
Fire 8	27 × 27 × 512		2	64	256	256
Maxpool8	13 × 12 × 512	3 × 3/2	0
Fire 9	13 × 13 × 512		2	64	256	256
Conv10	13 × 13 × 5	1 × 1/1 (×5)	1
Avgpool10	1 × 1 × 5	13 × 13/1	0