Monocular VO Based on Deep Siamese Convolutional Neural Network

Table 1

Configuration of the CNN in DSCNN.

Layer	Tensor size	Kernel size	Padding	Stride
Input		—	—	—
Conv1			3	2
Conv2			2	2
Conv3			2	2
Conv3_1			1	1
Conv4			1	2
Conv4_1			1	1
Conv5			1	2
Conv5_1			1	1
Conv6			1	2
Conv6_1			1	1
Max-pooling			0	2