Fig. 1. Feature vector of superpixel block

Fig. 2. Image properties. (a) Connected structure; (b) co-linearity structure; (c) co-planarity structure

Fig. 3. Structural diagram of GC-Net model

Fig. 4. Test results. (a) Left image of input binocular image; (b) depth estimation result; (c) true depth

Fig. 5. Classical learning algorithm architecture

Fig. 6. Overview flow chart of algorithm

Fig. 7. Overall flow chart based on composite view method

Fig. 8. Network architectures. (a) Depth estimation network; (b) pose estimation/explainability network