The paper introduces a novel approach, Adaptive Surface Normal (ASN), to estimate depth and surface normal from monocular images while incorporating geometric context. The ASN constraint dynamically determines reliable local geometry from randomly sampled candidates, ensuring that the predicted depth and surface normal are consistent with the geometric context. This method leverages geometric context to prioritize regions with significant geometric variations, improving the accuracy and robustness of the predicted normals. Extensive evaluations on indoor and outdoor datasets demonstrate the superiority of the proposed method over state-of-the-art techniques, showing its efficiency and robustness in generating high-quality 3D geometry from images. The code and data are available at https://www.xxlong.site/ASNDepth/.The paper introduces a novel approach, Adaptive Surface Normal (ASN), to estimate depth and surface normal from monocular images while incorporating geometric context. The ASN constraint dynamically determines reliable local geometry from randomly sampled candidates, ensuring that the predicted depth and surface normal are consistent with the geometric context. This method leverages geometric context to prioritize regions with significant geometric variations, improving the accuracy and robustness of the predicted normals. Extensive evaluations on indoor and outdoor datasets demonstrate the superiority of the proposed method over state-of-the-art techniques, showing its efficiency and robustness in generating high-quality 3D geometry from images. The code and data are available at https://www.xxlong.site/ASNDepth/.