Automatic Dataset Generation and Phase Unwrapping deep learning network Design for FPP

Fringe projection is widely used in scientific research and industrial measurement, due to its cost-effectiveness and precision. Nonetheless, acquiring massive fringe projection image datasets for real-world objects poses a challenge due to the limitation of photography. Furthermore, conventional method of developing deep learning models in Python often takes considerable effort in programming ancillary functions like image input, training progress control and so on. This paper introduces a practical methodology for researchers and engineers to effortlessly prepare datasets and design models for fringe projection using MATLAB. By applying algorithm to simulate fringe projection procedure, abundant image datasets can be generated and utilized for model pretraining. Besides, a phase unwrapping model was designed with MATLAB's deep network designer tool, which could achieve its expected function without complicated programming.