通过层选择性MLLMs实现细粒度人体姿态编辑评估
Fine-Grained Human Pose Editing Assessment via Layer-Selective MLLMs
AI总结：
本文提出HPE-Bench基准和基于层选择性MLLMs的统一框架，用于细粒度评估人体姿态编辑的真实性和质量。
摘要：
文本引导的人体姿态编辑在AIGC应用中取得了显著进展。然而，它仍然受到结构异常和生成伪影的困扰。现有评估指标往往将真实性检测与质量评估隔离，未能提供针对姿态特定不一致性的细粒度见解。为了解决这些限制，我们引入了HPE-Bench，一个包含来自17种最新编辑模型的1700个标准化样本的专用基准，提供真实性和多维质量评分。此外，我们提出了一种基于层选择性多模态大语言模型（MLLMs）的统一框架。通过使用对比性LoRA微调和一种新的层敏感性分析（LSA）机制，我们确定了最佳姿态评估特征层。我们的框架在真实性和多维质量回归方面均表现出色，有效弥合了取证检测与质量评估之间的差距。
英文摘要：
Text-guided human pose editing has gained significant traction in AIGC applications. However,it remains plagued by structural anomalies and generative artifacts. Existing evaluation metrics often isolate authenticity detection from quality assessment, failing to provide fine-grained insights into pose-specific inconsistencies. To address these limitations, we introduce HPE-Bench, a specialized benchmark comprising 1,700 standardized samples from 17 state-of-the-art editing models, offering both authenticity labels and multi-dimensional quality scores. Furthermore, we propose a unified framework based on layer-selective multimodal large language models (MLLMs). By employing contrastive LoRA tuning and a novel layer sensitivity analysis (LSA) mechanism, we identify the optimal feature layer for pose evaluation. Our framework achieves superior performance in both authenticity detection and multi-dimensional quality regression, effectively bridging the gap between forensic detection and quality assessment.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
