Table 2.

Network details of each node in Figure 2, including the convolutional kernel size of the reconstruction nodes, the input size, and the output size of feature maps.

Node	Kernel Size	Input Size	Output Size
0–0	-	H × W × 4	H × W × 8
0–1	3 × 3 × 16 × 16; 1 × 1 × 16 × 8	H × W × 16	H × W × 8
0–2	3 × 3 × 16 × 16; 1 × 1 × 16 × 8	H × W × 16	H × W × 8
0–3	3 × 3 × 16 × 16; 1 × 1 × 16 × 8	H × W × 16	H × W × 8
1–0	-	H × W × 4	H × W × 16
1–1	3 × 3 × 32 × 32; 1 × 1 × 32 × 16	H × W × 32	H × W × 16
1–2	3 × 3 × 32 × 32; 1 × 1 × 32 × 16	H × W × 32	H × W × 16
2–0	-	H × W × 4	H × W × 32
2–1	3 × 3 × 64 × 64; 1 × 1 × 64 × 32	H × W × 64	H × W × 32
3–0	-	H × W × 4	H × W × 64