Unsupervised Depth Estimation

Unsupervised depth estimation aims to infer 3D scene depth from only 2D images or video sequences, without relying on labeled depth data. Current research focuses on addressing challenges like handling dynamic scenes and color discrepancies between synthetic training data and real-world images, employing techniques such as temporal-spatial fusion, object-level depth estimation for moving objects, and novel network architectures like Swin Transformers and densely cascaded networks. These advancements improve the accuracy and robustness of depth estimation, with significant implications for applications such as autonomous driving, robotics, and 3D scene reconstruction.