Table 1.

The architecture of the proposed deep learning model.

Layer	Features (Train)	Features (Inference)	Kernel Size	Stride
Input	512 × 512 × 3	512 × 512 × 3	-	-
Padding	712 × 712 × 3	712 × 712 × 3	-	-
Conv1_1 + relu1_1	710 × 710 × 64	710 × 710 × 64	3 × 3	1
Conv1_2 + relu1_2	710 × 710 × 64	710 × 710 × 64	3 × 3	1
Pool1	355 × 355 × 64	355 × 355 × 64	2 × 2	2
Conv2_1 + relu2_1	355 × 355 × 128	355 × 355 × 128	3 × 3	1
Conv2_2 + relu2_2	355 × 355 × 128	355 × 355 × 128	3 × 3	1
Pool2	178 × 178 × 128	178 × 178 × 128	2 × 2	2
Conv3_1 + relu3_1	178 × 178 × 256	178 × 178 × 256	3 × 3	1
Conv3_2 + relu3_2	178 × 178 × 256	178 × 178 × 256	3 × 3	1
Conv3_3 + relu3_3	178 × 178 × 256	178 × 178 × 256	3 × 3	1
Pool3	89 × 89 × 256	89 × 89 × 256	2 × 2	2
Conv4_1 + relu4_1	89 × 89 × 512	89 × 89 × 512	3 × 3	1
Conv4_2 + relu4_2	89 × 89 × 512	89 × 89 × 512	3 × 3	1
Conv4_3 + relu4_3	89 × 89 × 512	89 × 89 × 512	3 × 3	1
Pool4	45 × 45 × 512	45 × 45 × 512	2 × 2	2
Conv5_1 + relu5_1	45 × 45 × 512	45 × 45 × 512	3 × 3	1
Conv5_2 + relu5_2	45 × 45 × 512	45 × 45 × 512	3 × 3	1
Conv5_3 + relu5_3	45 × 45 × 512	45 × 45 × 512	3 × 3	1
Pool5	23 × 23 × 512	23 × 23 × 512	2 × 2	2
Conv6 + relu6 + Drop6	17 × 17 × 4096	17 × 17 × 4096	7 × 7	1
Conv7 + relu7 + Drop7	17 × 17 × 4096	17 × 17 × 4096	1 × 1	1
Conv8	17 × 17 × 3	17 × 17 × 3	1 × 1	1
Deconv9	576 × 576 × 3	576 × 576 × 3	64 × 64	32
Cropping	512 × 512 × 3	512 × 512 × 3	-	-
Output Class Map	512 × 512 × 1	512 × 512 × 1	-	-