# Human Model

LHM (Large-scale Animatable Human Reconstruction Model) utilizes a multimodal transformer architecture for high-fidelity 3D head reconstruction, supporting the generation of animatable 3D human characters from a single image. The model can accurately preserve clothing geometry and texture, and is particularly excellent at restoring facial identity and details, making it suitable for application scenarios with high requirements for 3D reconstruction accuracy.

DressRecon is a method for reconstructing temporally consistent 4D human models from monocular video, focusing on handling very loose clothing or handheld object interactions. This technology combines general human prior knowledge learned from large-scale training data with specific 'skeletal bag' deformations tailored for individual videos through optimization during testing. DressRecon learns a neural implicit model to separate body and clothing deformations as distinct motion model layers. To capture subtle geometric features of clothing, it incorporates image-based prior knowledge, like human posture, surface normals, and optical flow, adjusted during the optimization process. The resulting neural field can be extracted into a temporally consistent mesh or optimized further into explicit 3D Gaussians to enhance rendering quality and achieve interactive visualization. DressRecon provides superior 3D reconstruction fidelity compared to previous technologies, even in datasets with highly challenging clothing deformations and object interactions.