ROMP provides a one-stage approach for extracting 3D human meshes and poses from single-camera images. The project includes multiple research works (ROMP, BEV, TRACE) for multi-person 3D reconstruction, depth estimation, and temporal tracking. It offers cross-platform APIs and supports exporting results to common 3D formats (FBX, GLB, BV). Built on PyTorch with SMPL body model integration.