Multi-Input Convolutional Neural Network for Flower Grading

Table 1

Architecture of models with different number of layers before merging.

Name of models	Convolutional layers
	C1	C2	C3	C4	C5	C6	FC1	FC2	ACC
M1	3 × 3,16		—	—	3 × 3,32		32	3
M2	3 × 3,16		—	—	3 × 3,32	3 × 3,32	32	3
M3	3 × 3,16	3 × 3,16	—	—	3 × 3,32	—	32	3
M4	3 × 3,16	3 × 3,16	—	—	5 × 5,64	5 × 5,32	32	3
M5	3 × 3,16	3 × 3,32	—	—	3 × 3,64	3 × 3,64	32	3
M6	5 × 5,16	5 × 5,16	—	—	5 × 5,32	5 × 5,32	32	3
M7	3 × 3,16	3 × 3,32	—	—	5 × 5,32	5 × 5,32	32	3
M8	3 × 3,16	3 × 3,32	3 × 3,32	—	3 × 3,64	—	64	3
M9	3 × 3,16	3 × 3,16	3 × 3,32	—	3 × 3,64	3 × 3,64	64	3
M10	5 × 5,16	3 × 3,16	3 × 3,32	3 × 3,32	3 × 3,32	—	32	3
M11	5 × 5,16	5 × 5,16	3 × 3,32	3 × 3,32	3 × 3,32	3 × 3,32	32	3

represents the convolutional layers in each branch before merging. Each branch has only one convolutional layer. “3 × 3,32” represents the size of filter kernels which is 3 × 3 and the number of kernels is 32. All the strides of kernels are set to 1 × 1.