NeuS is a novel neural surface reconstruction method that uses volume rendering to learn a neural signed distance function (SDF) representation for multi-view reconstruction. Unlike existing methods that require foreground masks or struggle with self-occlusion and thin structures, NeuS represents surfaces as the zero-level set of an SDF and employs a new volume rendering method to train the SDF without mask supervision. This approach avoids inherent geometric errors in surface reconstruction by ensuring unbiased surface representation in the first-order approximation of SDF. Experiments on the DTU and BlendedMVS datasets show that NeuS outperforms state-of-the-art methods like IDR and NeRF in terms of reconstruction quality, especially for complex structures and self-occluded objects. The method introduces a novel weight function that is both unbiased and occlusion-aware, enabling accurate surface reconstruction without the need for masks. NeuS also handles thin structures effectively, demonstrating superior performance in reconstructing challenging scenes with abrupt depth changes. The approach is robust to varying lighting conditions and can reconstruct high-quality surfaces from 2D images. However, it has limitations in handling textureless objects and relies on a single scale parameter for probability distribution modeling. Overall, NeuS provides a powerful framework for multi-view surface reconstruction with high fidelity and robustness.NeuS is a novel neural surface reconstruction method that uses volume rendering to learn a neural signed distance function (SDF) representation for multi-view reconstruction. Unlike existing methods that require foreground masks or struggle with self-occlusion and thin structures, NeuS represents surfaces as the zero-level set of an SDF and employs a new volume rendering method to train the SDF without mask supervision. This approach avoids inherent geometric errors in surface reconstruction by ensuring unbiased surface representation in the first-order approximation of SDF. Experiments on the DTU and BlendedMVS datasets show that NeuS outperforms state-of-the-art methods like IDR and NeRF in terms of reconstruction quality, especially for complex structures and self-occluded objects. The method introduces a novel weight function that is both unbiased and occlusion-aware, enabling accurate surface reconstruction without the need for masks. NeuS also handles thin structures effectively, demonstrating superior performance in reconstructing challenging scenes with abrupt depth changes. The approach is robust to varying lighting conditions and can reconstruct high-quality surfaces from 2D images. However, it has limitations in handling textureless objects and relies on a single scale parameter for probability distribution modeling. Overall, NeuS provides a powerful framework for multi-view surface reconstruction with high fidelity and robustness.