LUMA：用轻量通用掩码适配器统一评测图像分割骨干网络

View PDF HTML (experimental)

Abstract:Comparing transformer backbones for image segmentation is confounded: each is paired with a different decoder, recipe, and pretraining, so reported differences rarely reflect the backbone itself. We introduce the Lightweight Universal Mask Adapter (LUMA), a lightweight, backbone-agnostic mask-transformer head that treats any backbone as a black-box feature extractor, letting a set of queries read from its features through cheap cross-attention. LUMA matches the accuracy of EoMT, the state-of-the-art efficient ViT-segmenter, at cost parity, while attaching unchanged to hierarchical, convolutional, and mixture-of-experts backbones alike. Holding this head fixed, we benchmark 20 backbones, 11 pretraining schemes and a range of resolutions on ADE20K, COCO, and Cityscapes under one modern recipe. We find that ``efficient'' token mixers fail to deliver efficiency even at the high resolutions that motivate them, with plain ViT holding the throughput Pareto front at every resolution. Additionally, the pretraining objective, not the architecture, the lever the field has tuned hardest, governs segmentation quality.
Code: this https URL

Comments:	Accepted at WACV 2027. v2: Camera-ready version with supplementary material included
Subjects:	Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Performance (cs.PF)
MSC classes:	68T45
ACM classes:	I.2.10; I.5.0
Cite as:	arXiv:2607.00687 [cs.CV]
	(or arXiv:2607.00687v2 [cs.CV] for this version)
	https://doi.org/10.48550/arXiv.2607.00687 arXiv-issued DOI via DataCite

Submission history

From: Tobias Nauen [view email]
[v1] Wed, 1 Jul 2026 09:35:24 UTC (498 KB)
[v2] Mon, 26 Oct 2026 08:40:15 UTC (1,014 KB)