Depth Restoration

Depth restoration aims to improve the accuracy and completeness of depth maps generated by various sensors, addressing limitations like noise, missing data, and inaccuracies caused by challenging materials (e.g., transparency, specular reflection). Current research focuses on developing self-supervised learning methods and neural network architectures, often incorporating data from multiple sensors or leveraging polarization information to enhance depth estimation. These advancements are crucial for improving the performance of robotics, augmented/mixed reality, and 3D mapping applications that rely on accurate depth perception.