Monocular VO Based on Deep Siamese Convolutional Neural Network

Figure 3

Relative poses of consecutive image frames.