ICON (Implicit Clothed humans Obtained from Normals) is a computer vision research project that reconstructs detailed 3D human meshes with clothing from single-view images. It leverages implicit neural functions combined with normal map estimation and SMPL/SMPLX body models to generate animatable avatars. The method uses PyTorch and PyTorch Lightning for training and inference, and is designed for applications in computer graphics, virtual humans, and the metaverse.