E-Commerce Picture Text Recognition Information System Based on Deep Learning

Table 1

FPN-NET v2 backbone network structural parameters.

Group	Block	Layer	Output
	Input		512 × 512 × 3
Conv1_x	Initializer block		128 × 128 × 32
conv2_x	MWI-dense block 1	Layer × 6	64 × 64 × 64
	Transition layer	1 × 1 conv; pooling; SE module
conv3_x	MWI-dense block 2	Layer × 12	32 × 32 × 256
	Transition layer	1 × 1 conv; pooling; SE module
con\4_x	MWl-dense block 3	layer × 24	I6 × 16 × 256
	Transition layer	1 × 1 conv; pooling; SE module
conv5_x	MWI-dense block 4	layer × 8	8 × 8 × 256
	Transition layer	1 × 1 conv; SE module
conv6_x	conv6_l	conv3 × 3,s = 2	4 × 4 × 512
	conv6_2	couv3 × 3,s = 2	2 × 2 × 512
	conv6_2	conv3 × 3,s = 2	1 × 1 × 512