arXiv：cs.LG（机器学习，全量分类）· Mingzhi Xu, Tao Zhou, Yong Li, Yizhe Zhang·· 14 小时前AI 评分37
Spatial Lifting：面向密集预测任务的高效建模方法
Spatial Lifting for Dense Prediction
AI 导读
研究者提出 Spatial Lifting（SL），将 2D 图像等标准输入提升到更高维空间，再用 3D U-Net 等对应高维网络处理，在基准任务上取得良好性能的同时降低推理成本并大幅减少模型参数量。SL 在提升维度上产生内在结构化输出，支持训练时的密集监督，并在测试时实现单次前向传播的自一致性质量与不确定性估计。
正文
Abstract:We present Spatial Lifting (SL), a novel methodology for dense prediction tasks. SL operates by lifting standard inputs, such as 2D images, into a higher-dimensional space and subsequently processing them using networks designed for that higher dimension, such as a 3D U-Net. Counterintuitively, this dimensionality lifting allows us to achieve good performance on benchmark tasks compared to conventional approaches, while reducing inference costs and \textbf{drastically lowering the number of model parameters}. The SL framework produces intrinsically structured outputs along the lifted dimension. This emergent structure facilitates dense supervision during training and enables single-forward-pass self-consistency-based quality and uncertainty estimation at test time. Spatial Lifting introduces a simple and general modeling strategy that offers a promising path toward more efficient, accurate, and reliable deep networks for dense prediction tasks in vision.
| Comments: | 28 pages 5 figures |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.00017 [cs.CV] |
| (or arXiv:2610.00017v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.00017 arXiv-issued DOI via DataCite |
Submission history
From: Mingzhi Xu [view email]
[v1]
Mon, 27 Jul 2026 15:22:14 UTC (10,029 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org