Yolo-Based Improvements in Remote Sensing Image Applications

Table 1

The detailed structure of SGEDarknet53 where K, S, and represent the kernel size, stride, and padding, respectively.

Layers	Filters	Size	OutPut size
Convolutional	32	K = 3, S = 1,  = 1	416 × 416 × 32
Convolutional	64	K = 3, S = 2,  = 1	208 × 208 × 64
Convolutional	32	K = 1, S = 1,  = 0	208 × 208 × 32
SGEResidual	64	K = 3, S = 1,  = 1	208 × 208 × 64
Convolutional	128	K = 3, S = 2,  = 1	104 × 104 × 128
2×Convolutional	64	K = 1, S = 1,  = 0	104 × 104 × 64
2×SGEResidual	128	K = 3, S = 1,  = 1	104 × 104 × 128
Convolutional	256	K = 3, S = 2,  = 1	52 × 52 × 256
8×Convolutional	128	K = 1, S = 1,  = 0	52 × 52 × 128
8×SGEResidual	256	K = 3, S = 1,  = 1	52 × 52 × 256
Convolutional	512	K = 3, S = 2,  = 1	26 × 26 × 512
8×Convolutional	256	K = 1, S = 1,  = 0	26 × 26 × 256
8×SGEReSidual	512	K = 3, S = 1,  = 1	26 × 26 × 512
Convolutional	1024	K = 3, S = 2,  = 1	13 × 13 × 1024
4×Convolutional	512	K = 1, S = 1,  = 0	13 × 13 × 512
4×SGEReSidual	1024	K = 3, S = 1,  = 1	13 × 13 × 1024