DARA：领域感知与关系感知适配器实现视觉定位的高效参数微调
DARA: Domain- and Relation-aware Adapters Make Parameter-efficient Tuning for Visual Grounding
AI总结：
本文提出 DARA 方法，通过领域感知适配器和关系感知适配器实现视觉定位的参数高效迁移学习，仅微调 2.13% 的骨干参数即在三个基准上平均精度提升 0.81%。
摘要：
视觉定位（Visual Grounding, VG）是一项具有挑战性的任务，旨在根据文本描述在图像中定位目标对象。近年来，VG 模型规模的迅速扩大显著提升了性能，但也给微调过程中的计算成本带来了沉重负担。本文探索了应用参数高效迁移学习（PETL）来高效地将预训练的视觉-语言知识迁移到 VG 任务中。具体而言，我们提出了 DARA，一种新颖的 PETL 方法，包含领域感知适配器（Domain-aware Adapters, DA Adapters）和关系感知适配器（Relation-aware Adapters, RA Adapters），用于视觉定位。DA Adapters 首先将模态内表示迁移为更适合 VG 领域的细粒度表示。随后，RA Adapters 通过共享权重来桥接两种模态之间的关系，从而提升空间推理能力。在广泛使用的基准数据集上的实验结果表明，与全量微调和其他 PETL 方法相比，DARA 在节省大量更新参数的同时取得了最佳精度。值得注意的是，仅使用 2.13% 的可调骨干网络参数，DARA 在三个基准数据集上的平均精度相较于基线模型提升了 0.81%。我们的代码已开源，地址为 https://github.com/liuting20/DARA。
英文摘要：
Visual grounding (VG) is a challenging task to localize an object in an image based on a textual description. Recent surge in the scale of VG models has substantially improved performance, but also introduced a significant burden on computational costs during fine-tuning. In this paper, we explore applying parameter-efficient transfer learning (PETL) to efficiently transfer the pre-trained vision-language knowledge to VG. Specifically, we propose \textbf{DARA}, a novel PETL method comprising \underline{\textbf{D}}omain-aware \underline{\textbf{A}}dapters (DA Adapters) and \underline{\textbf{R}}elation-aware \underline{\textbf{A}}dapters (RA Adapters) for VG. DA Adapters first transfer intra-modality representations to be more fine-grained for the VG domain. Then RA Adapters share weights to bridge the relation between two modalities, improving spatial reasoning. Empirical results on widely-used benchmarks demonstrate that DARA achieves the best accuracy while saving numerous updated parameters compared to the full fine-tuning and other PETL methods. Notably, with only \textbf{2.13\%} tunable backbone parameters, DARA improves average accuracy by \textbf{0.81\%} across the three benchmarks compared to the baseline model. Our code is available at \url{https://github.com/liuting20/DARA}.
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。
