Fig. 1. DCSHF. (a) 3D schematic diagram; (b) cross-section micrograph

Fig. 2. Experimental system of optical fiber specklegram. (a) 3D schematic(inset is SMF-MMF-DCSHF sensing structure); (b) 2D side view

Fig. 3. Logic diagram of automated experiment

Fig. 4. Schematic diagram of curvature auto-determination device (inset is detection result of scale alignment algorithm)

Fig. 5. Structure of VggNet 16

Fig. 6. Light speckle passes through feature maps of each layer of model

Fig. 7. Bending prediction of SMF-MMF structure based on VggNet 16 (classified)

Fig. 8. Schematic diagram of curvature calculation

Fig. 9. Curvature prediction of SMF-MMF structure based on VggNet 16 (classified)

Fig. 10. Loss trend chart of VggNet 16 (regression)

Fig. 11. Curvature prediction of SMF-MMF-DCSHF structure based on VggNet 16 (regression)

Fig. 12. Comparison of SMF-MMF light speckle

Fig. 13. Comparison of SMF-MMF-DCSHF light speckle

Fig. 14. Comparison of similarity parameters for SMF-MMF light speckle with different algorithms

Fig. 15. Comparison of similarity parameters for SMF-MMF-DCSHF light speckle with different algorithms

Fig. 16. Comparison of light speckle for two structures with different algorithms

Fig. 17. Curvature prediction of SMF-MMF structure based on VggNet 16 (regression)

Fig. 18. Curvature prediction of SMF-MMF-DCSHF structure based on ResNet 50 (regression)

Fig. 19. Analysis of activation area of model