GaussianBody is a novel method for reconstructing clothed human models from monocular RGB videos using 3D Gaussian Splatting. The method addresses the challenges of dynamic human reconstruction by incorporating explicit pose-guided deformation to associate Gaussians across canonical and observation spaces. A physically-based prior with regularized transformations helps mitigate ambiguity between the two spaces. During training, a pose refinement strategy updates the pose regression to compensate for inaccurate initial estimates, and a split-with-scale mechanism enhances the density of regressed point clouds. The method achieves state-of-the-art photorealistic novel-view rendering results with high-quality details for dynamic clothed human bodies, along with explicit geometry reconstruction. The method is evaluated on monocular videos of dynamic clothed humans, showing superior reconstruction quality in rendering details and geometry recovery, while requiring much less training time (approximately one hour) and almost real-time rendering speed. Ablation studies validate the effectiveness of each component in the method. GaussianBody extends the 3D-GS representation to clothed human reconstruction by utilizing an articulated human model for guidance. Specifically, forward linear blend skinning (LBS) is used to deform Gaussians from the canonical space to each observation space per frame. A physically-based prior is optimized for Gaussians in the observation space to mitigate overfitting. Local-rigidity, local-rotation, and local-isometry losses are introduced to maintain the local geometry property of the deformed 3D Gaussians. A split-with-scale strategy enhances point cloud density, and a pose refinement approach addresses texture blurring issues. The method achieves high-quality results and fast rendering by incorporating 3D-GS. It is evaluated on the PeopleSnapshot dataset, showing significant improvements in PSNR and SSIM metrics. The method outperforms other approaches in capturing detailed reconstructions, particularly in cloth textures and human body details. It also demonstrates robustness in handling dynamic scenes and inaccurate pose parameters. The method is compared with other baselines and state-of-the-art works, showing competitive performance, relatively fast training speeds, and the capability to train with higher resolution images.GaussianBody is a novel method for reconstructing clothed human models from monocular RGB videos using 3D Gaussian Splatting. The method addresses the challenges of dynamic human reconstruction by incorporating explicit pose-guided deformation to associate Gaussians across canonical and observation spaces. A physically-based prior with regularized transformations helps mitigate ambiguity between the two spaces. During training, a pose refinement strategy updates the pose regression to compensate for inaccurate initial estimates, and a split-with-scale mechanism enhances the density of regressed point clouds. The method achieves state-of-the-art photorealistic novel-view rendering results with high-quality details for dynamic clothed human bodies, along with explicit geometry reconstruction. The method is evaluated on monocular videos of dynamic clothed humans, showing superior reconstruction quality in rendering details and geometry recovery, while requiring much less training time (approximately one hour) and almost real-time rendering speed. Ablation studies validate the effectiveness of each component in the method. GaussianBody extends the 3D-GS representation to clothed human reconstruction by utilizing an articulated human model for guidance. Specifically, forward linear blend skinning (LBS) is used to deform Gaussians from the canonical space to each observation space per frame. A physically-based prior is optimized for Gaussians in the observation space to mitigate overfitting. Local-rigidity, local-rotation, and local-isometry losses are introduced to maintain the local geometry property of the deformed 3D Gaussians. A split-with-scale strategy enhances point cloud density, and a pose refinement approach addresses texture blurring issues. The method achieves high-quality results and fast rendering by incorporating 3D-GS. It is evaluated on the PeopleSnapshot dataset, showing significant improvements in PSNR and SSIM metrics. The method outperforms other approaches in capturing detailed reconstructions, particularly in cloth textures and human body details. It also demonstrates robustness in handling dynamic scenes and inaccurate pose parameters. The method is compared with other baselines and state-of-the-art works, showing competitive performance, relatively fast training speeds, and the capability to train with higher resolution images.