DL-SL3D: Open-source implementation for learning-enabled single-shot structured-light 3D imaging

This repository provides the open-source implementation of DL-SL3D, a learning-enabled single-shot structured-light 3D imaging framework. The released code includes the training, deployment, and real-time inference pipeline of the lightweight phase retrieval network (LPR-Net), covering PyTorch training, ONNX export, TensorRT optimization, and GPU-accelerated inference. Representative fringe images and example outputs are provided to facilitate reproducibility. System calibration and hardware-specific configurations follow standard structured-light metrology practices and are described in the accompanying manuscript.