ElasticDino：利用弹性形变生成稠密语义特征
ElasticDino: Dense Semantic Features Using Elastic Deformations
AI 导读
arXiv 论文提出 ElasticDino，一个轻量级适配器，通过空间弹性形变为 ViT 生成与输入图像像素对齐的高分辨率特征图，解决自注意力二次复杂度导致的低分辨率特征与空间伪影问题。该方法用线性头在语义分割和深度估计等下游稠密任务上训练，无需 ground truth，在多个流行稠密任务基准上以更好的计算性能取得更高精度。
正文 · 原文
Abstract:Vision transformers (ViTs) serve as the backbone for countless vision tasks, including dense prediction tasks, such as segmentation and depth estimation, even though the quadratic complexity of self-attention severely limits feature resolution and induces spatial artifacts. In this work, we introduce ElasticDino, a lightweight adapter which produces high-resolution feature maps, adhering to the fine details of the input image. As a key insight, we observe the consistency of artifacts across channels, and simultaneously resolve them to match the input image structure. Using a slim architecture based on spatial deformations, ElasticDino produces pixel-aligned, high-resolution feature maps efficiently. Leveraging deformations allows strictly preserving sharp semantics, and thereby avoiding the smoothness biases inherent to alternative approaches. Due to the absence of ground truth, we employ pixel-level semantic segmentation and depth estimation with linear heads to predict a set of downstream dense tasks, guiding the deformation process without leaking task-specific semantics into the model. We demonstrate superior accuracy over current methods on a range of popular dense task benchmarks, with better computational performance.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.16984 [cs.CV] |
| (or arXiv:2611.16984v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2611.16984 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ulysse Mizrahi [view email]
[v1]
Thu, 12 Nov 2026 15:50:04 UTC (9,294 KB)
来源：arXiv cs.AI · arxiv.org