ZipDepth：在任何设备上随时随地实现轻量级零样本单目深度估计
ZipDepth: Bringing Lightweight Zero-Shot Monocular Depth Anywhere, on Any Device
AI总结：
研究单目深度估计难题，提出ZipDepth轻量级网络，结合可重新参数化编码器-解码器与知识蒸馏，在多域训练集训练，参数仅610万，能在多设备实时运行，在五个基准测试中平衡零样本精度与部署效率，向高精度基础模型迈进。
摘要：
单目深度估计通过基础模型取得了显著进展，实现了强大的零样本泛化，但计算需求使其难以应用于嵌入式和移动平台。现有的轻量级替代方案几乎都是在单域、自监督范式下开发的，在域转移时表现不佳。我们提出了ZipDepth，这是一个紧凑的单目深度网络，通过将高效的可重新参数化编码器-解码器与来自基础模型在大型多域训练集上的大规模知识蒸馏相结合来弥合这一差距。ZipDepth仅包含610万个参数，从服务器GPU到功率受限设备都能实时运行，在五个基准测试中，在轻量级模型中实现了零样本精度和部署效率之间的最佳权衡，朝着参数多50倍的基础模型的精度迈出了重要一步。
英文摘要：
Monocular depth estimation has seen remarkable progress through foundation models achieving robust zero-shot generalization, yet their computational demands place them far beyond the reach of embedded and mobile platforms. Lightweight alternatives exist, but have been developed almost exclusively within single-domain, self-supervised paradigms, failing silently under domain shift. We present ZipDepth, a compact monocular depth network that bridges this gap by combining an efficient reparameterizable encoder-decoder with large-scale knowledge distillation from a foundation model over a large multi-domain training set. Comprising just 6.1M parameters, ZipDepth runs at real-time rates from server GPUs to power-constrained devices, achieving the best trade-off between zero-shot accuracy and deployment efficiency among lightweight models across five benchmarks, taking a significant step towards the accuracy of foundation models with 50x more parameters.
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。
