CORGI: Consistency-Aware 3D Dog Reconstruction from a Single Image in the Wild

2026-06-30

AI Summary

Reconstructing high-fidelity, animatable 3D dog models from a single in-the-wild image is challenging due to extreme pose variations, the absence of 3D supervision, and lack of multi-view data. This work proposes CORGI, a novel framework that introduces Canonical-Driven Orbital Generation to synthesize reliable multi-view observations. It integrates a consistency-aware deformable 3D Gaussian Splatting (CA-3DGS) module with a deformation-conditioned generative inpainting (DCGR) component, enabling high-quality 3D reconstruction without any 3D supervision. The method achieves state-of-the-art performance across diverse dog breeds, producing geometrically accurate, visually consistent, and readily animatable 3D assets suitable for downstream applications.

Abstract

Reconstructing high-fidelity 3D models of highly articulated animals, such as dogs, from a single in-the-wild image remains a formidable challenge. In this paper, we introduce CORGI, a novel framework for consistency-aware 3D dog reconstruction from a single unconstrained image that completely eliminates the need for 3D supervision. To overcome generative inconsistencies and the lack of multi-view capture, our pipeline introduces three core components. First, we propose a Canonical-Driven Orbital Generation (CDOG) strategy, utilizing specialized Canonical and Orbit LoRAs to normalize arbitrary input poses and synthesize reliable 360-degree video observations. Second, we design a Consistency-aware Deformable 3DGS (CA-3DGS) module that anchors on a D-SMAL prior, explicitly modeling per-view generative errors through dedicated neural deformation fields to learn accurate vertex-level displacements. Finally, to eliminate structural distortions and recover high-frequency details, we introduce a self-supervised Deformation-Conditioned Generative Repair (DCGR) module. Extensive experiments demonstrate that CORGI achieves state-of-the-art performance, generalizing seamlessly across diverse dog breeds to produce geometrically accurate, visually coherent, and fully animatable 3D assets ready for downstream applications.

Problem

• 3D reconstruction
• single-image
• articulated animals
• in-the-wild
• dog

Innovation

• 3D reconstruction
• single-image
• consistency-aware
• LoRA
• self-supervised