natowi/3D-Reconstruction-with-Deep-Learning-Methods

Curated catalog of deep learning approaches for 3D reconstruction spanning monocular depth estimation, multi-view stereo, point cloud processing, and semantic scene understanding. Projects leverage frameworks like PyTorch, TensorFlow, and cuDNN, with techniques including CNNs, GANs, self-supervised learning, and point cloud networks (PointNet). Covers end-to-end pipelines from synthetic data generation (BlenderProc) through registration, odometry, and mesh reconstruction for robotics and computer vision applications.

1,010 stars. No commits in the last 6 months.