Generator-Refiner-Examiner: A Tri-Module Data Augmentation Framework for 3D Human Avatar Learning from Monocular Videos

2026-05-22

AI Summary

This work addresses the challenge of detail loss in monocular video-based 3D human avatar reconstruction due to limited training frames. To overcome this limitation, the authors propose TrioMan, a novel framework that introduces, for the first time, a collaborative data augmentation mechanism comprising three modules: a generator, an optimizer, and a validator. The approach synthesizes diverse training samples by perturbing Gaussian poses and camera parameters, refines geometry and appearance via a one-step diffusion model, and employs a dual-branch attention-based consistency evaluator to select high-quality augmented data. Experiments on the X-Humans and NeuMan benchmarks demonstrate that TrioMan substantially outperforms existing methods, achieving significant improvements in both reconstruction quality and fine-detail fidelity.

Abstract

This paper addresses the challenge of reconstructing photorealistic and animatable 3D human avatars from monocular videos. While existing methods rely on combining per-subject optimization with generic human priors, they often fail to capture fine-grained details when training frames are limited. To mitigate this data scarcity, we propose TrioMan, a systematic tri-module framework for augmented 3D avatar learning. Our approach comprises three synergistic components. The Generator creates diverse unseen samples by imposing Gaussian perturbations on pose and camera. The Refiner improves the quality of generated data through one-step diffusion guided by texture and geometry cues. The Examiner selects subject-consistent samples using a dual-branch attention-based similarity evaluation. Experiments on the X-Humans and NeuMan benchmarks show that TrioMan outperforms state-of-the-art methods.

Problem

• 3D human avatar
• monocular video
• data scarcity
• photorealistic reconstruction
• animatable avatar

Innovation

• data augmentation
• 3D human avatar
• monocular video
• diffusion refinement
• attention-based selection