The paper introduces a Deep Ordinal Regression Network (DORN) for monocular depth estimation, addressing the ill-posed nature of the problem by discretizing depth values and training a regression network using an ordinal regression loss. The method aims to improve convergence and accuracy compared to traditional mean squared error (MSE) regression, which often leads to slow convergence and suboptimal solutions. DORN employs a spacing-increasing discretization (SID) strategy to better capture the ordering of depth values, avoiding the need for complex multi-scale networks or skip-connections. The network architecture includes a dense feature extractor, a multi-scale feature learner, and a full-image encoder to capture global context. Experiments on four challenging datasets (KITTI, ScanNet, Make3D, and NYU Depth v2) demonstrate that DORN achieves state-of-the-art performance, outperforming previous methods by significant margins. The code for DORN is available at <https://github.com/hufu6371/DORN>.The paper introduces a Deep Ordinal Regression Network (DORN) for monocular depth estimation, addressing the ill-posed nature of the problem by discretizing depth values and training a regression network using an ordinal regression loss. The method aims to improve convergence and accuracy compared to traditional mean squared error (MSE) regression, which often leads to slow convergence and suboptimal solutions. DORN employs a spacing-increasing discretization (SID) strategy to better capture the ordering of depth values, avoiding the need for complex multi-scale networks or skip-connections. The network architecture includes a dense feature extractor, a multi-scale feature learner, and a full-image encoder to capture global context. Experiments on four challenging datasets (KITTI, ScanNet, Make3D, and NYU Depth v2) demonstrate that DORN achieves state-of-the-art performance, outperforming previous methods by significant margins. The code for DORN is available at <https://github.com/hufu6371/DORN>.