Weakly Supervised Deep Semantic Segmentation Using CNN and ELM with Semantic Candidate Regions

Table 1

Network structure of CNN.

structure	input	operate			output
		convolution	Nonlinear mapping	Pooling
Conv1	27×27×3	64×3×3 ×3 stride 1	ReLU		27×27×64
Conv2	27×27×64	256×5×5×64 stride 1	ReLU	2×2 pool	13×13×256
Conv3	13×13×256	256×3×3×256 stride 1	ReLU		13×13×256
Conv4	13×13×256	256×3×3×256 stride 1	ReLU		13×13×256
Conv5	13×13×256	512×3×3×256 stride 1	ReLU	2×2 pool	6×6×512
Fc6	6×6×512	4096	ReLU		4096
Fc7	4096	4096	ReLU		4096
Fc8	1000	1000			1000