RecycleLoRA: Rank-Revealing QR-Based Dual-LoRA Subspace Adaptation for Domain Generalized Semantic Segmentation

Chanseul Cho; Seokju Yun; Jeaseong Jeon; Seungjae Moon; Youngmin Ro

cs.CVMar 30, 2026

RecycleLoRA: Rank-Revealing QR-Based Dual-LoRA Subspace Adaptation for Domain Generalized Semantic Segmentation

Authors: Chanseul Cho, Seokju Yun, Jeaseong Jeon, Seungjae Moon, Youngmin Ro

Organizations: Machine Intelligence Laboratory, University of Seoul, Korea

Abstract

Domain Generalized Semantic Segmentation (DGSS) aims to maintain robust performance across unseen target domains. Vision Foundation Models (VFMs) offer rich multi-domain knowledge that can enhance generalization. However, strategies for actively exploiting the rich subspace structures within VFMs remain under-explored, with many existing methods focusing primarily on preserving pre-trained knowledge. Furthermore, their LoRA components often suffer from limited representational diversity and inefficient parameter utilization. We propose RecycleLoRA, which addresses both challenges by employing Rank-Revealing QR Decomposition (RRQR) to systematically exploit VFM's subspace structures and enhance LoRA's representational richness. Our main adapter leverages minor subspace directions identified by RRQR to learn diverse and independent features, achieving competitive performance even when used alone. We further introduce a sub adapter that carefully refines major directions with minimal adjustments, providing complementary improvements to the main adapter's strong baseline performance. This design enables the dual adapters to learn distinct representations without requiring additional regularization losses. Our systematic exploitation of pre-trained subspace structures through RRQR-based initialization leads to superior domain generalization performance. RecycleLoRA achieves state-of-the-art performance on both synthetic-to-real generalization and real-to-real generalization tasks without complex architectures or additional inference latency.

Figures & tables

Figure 1: Comparison of synthetic-to-real generalization performance (mIoU,%) between our proposed RecycleLoRA and the previous SOTA method, SoMA.

Figure 2: RecycleLoRA Framework Overview. This figure illustrates the overall workflow of RecycleLoRA. (a) Rank-Revealing QR Decomposition (RRQR) is applied to the pre-trained weight matrix to identify subspace directions ranked by importance. (b) Among the recyclable subspaces selected through RRQR, the minor directions are assigned as initialization values for the main adapter, while the major directions are assigned to the sub adapter. (c) The main adapter’s B matrix is initialized with the minor directions, and its A matrix is sparsely initialized by mapping these directions to their corresponding column indices. (d) The sub adapter’s B matrix is initialized with the major directions, and its A matrix is sparsely initialized by mapping these directions to their corresponding column indices.

LoRA Column‑wise Norm Statistics
	Mean $\,\ell_{2}$ norm	Mean $\,\ell_{2}$ norm	Average	Maximum
	(selected)	(non‑selected)	ratio	ratio
All layers	0.138245	0.113566	1.22 $\times$	1.63 $\times$

Table 1: Comparison of

\ell_{2}

-norm statistics between selected and non-selected columns in LoRA matrix A after training.

Effective Rank & Rank‑Efficiency Statistics
Target rank	RecycleLoRA (Ours)		SoMA
Target rank	Effective rank	Rank efficiency	Effective rank	Rank efficiency
16	13.60	0.850	9.78	0.611
32	24.65	0.770	20.80	0.650

Table 2: Effective rank and rank efficiency comparison. Rank efficiency is calculated as the ratio of effective rank to target rank, measuring parameter utilization. RecycleLoRA consistently achieves higher efficiency than SoMA across different rank settings.

Figure 3: Cosine similarity heatmaps of LoRA components for (a) RecycleLoRA and (b) SoMA at different ranks (r=16, 32). Left: pairwise similarity among rows of $\mathbf{A}$ . Right: pairwise similarity among columns of $\mathbf{B}$ . Darker blue colors represent lower similarity.

Figure 4: Block-wise subspace similarity $\phi$ between main and sub adapters measured using Grassmann Distance on the low-rank matrices. Lower values indicate more orthogonal subspaces.

Figure 5: Visualization of adapter-induced feature modifications via PCA projection. (a) Input. (b) main adapter PCA Visualization. (c) sub adapter PCA Visualization. The divergent activation patterns reveal complementary feature learning between the dual adapters.

Synthetic-to-Real Generalization
Method	Venue	Backbone	Trained on GTAV
Method	Venue	Backbone	$\rightarrow$ Citys.	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
CLOUDS [ 4 ]	CVPR2024	CLIP-CN-L	60.20	57.40	67.00	61.50
VLTSeg [ 33 ]	ACCV2024	EVA02-L	65.30	58.30	66.00	63.20
DoRA [ 49 ]	ICML2024	DINOv2-L	66.12	59.31	67.07	64.17
VPT [ 35 ]	ECCV2022	DINOv2-L	68.75	58.64	68.32	65.24

Table 3: Domain generalization results (mIoU %) under the synthetic-to-real setting. Our method is highlighted in gray . Bold and underlined indicate best and second-best results.

Real-to-Real Generalization
Method	Venue	Backbone	Trained on Cityscapes
Method	Venue	Backbone	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
HGFormer [ 17 ]	CVPR2023	Swin-L	61.50	72.10	66.80
CMFormer [ 6 ]	AAAI2024	Swin-L	62.60	73.60	68.10
PDAF [ 9 ]	ICCV2025	Swin-L	63.00	74.10	68.55
SET [ 74 ]	TIP2021	DINOv2-L	65.07	75.67	70.37

Table 4: Domain generalization results (mIoU %) under the real-to-real setting. Our method is highlighted in gray . Bold and underlined indicate best and second-best results.

Main	Sub	Params.	$\rightarrow$ Citys.	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
	✔	1.6M	70.64	60.56	71.11	67.44
✔		12.6M	72.92	61.22	71.75	68.63
✔	✔	14.2M	73.01 ( $\uparrow$ 0.09)	61.77 ( $\uparrow$ 0.55)	72.07 ( $\uparrow$ 0.32)	68.95 ( $\uparrow$ 0.32)

Table 5: Ablation study on the Main and Sub Adapter components of RecycleLoRA. Best results in bold .

Method	lr.	$\rightarrow$ Citys.	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
Main Adapter	1e-4	72.92	61.22	71.75	68.63
Main Adapter	5e-5	69.46 ( $\downarrow$ 3.46)	62.22 ( $\uparrow$ 1.00)	68.23 ( $\downarrow$ 3.52)	66.64 ( $\downarrow$ 1.99)
LoRA	1e-4	70.36	60.21	69.95	66.84
LoRA	5e-5	69.40 ( $\downarrow$ 0.96)	59.62 ( $\downarrow$ 0.59)	69.20 ( $\downarrow$ 0.75)	66.07 ( $\downarrow$ 0.77)
Sub Adapter	1e-4	68.60	60.74	67.54	65.63
Sub Adapter	5e-5	70.64 ( $\uparrow$ 2.04)	60.56 ( $\downarrow$ 0.18)	71.11 ( $\uparrow$ 3.57)	67.44 ( $\uparrow$ 1.81)

Table 6: Comparative analysis of learning rate sensitivity across initialization strategies. Standard LoRA employs Kaiming initialization while RecycleLoRA’s Sub Adapter utilizes RRQR top-ranked directions. The best result for each method is shown in bold .

Synthetic-to-Real Generalization
Method	Venue	Backbone	Trained on GTAV
Method	Venue	Backbone	$\rightarrow$ Citys.	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
Rein [ 71 ]	CVPR2024	EVA02-L	65.30	60.50	64.90	63.60
FADA [ 5 ]	NeurIPS2024	EVA02-L	66.70	61.90	66.10	64.90
DepthForge [ 11 ]	ICCV2025	EVA02-L	68.00	61.70	67.50	65.73
SoMA [ 76 ]	CVPR2025	EVA02-L	68.05	60.81	68.33	65.73

Table 7: Domain generalization performance (mIoU, %) on the EVA02-L backbone under the synthetic-to-real setting. Bold and underlined indicate the best and second-best results, respectively.

Hyperparameter	Synthetic-to-Real	Real-to-Real
backbone	DINOv2-L	DINOv2-L
main rank	32	32
sub rank	4	2
main lr mult.	1.0	1.0
sub lr mult.	0.5	0.5
learning rate	1e-4	1e-4

Table 8: Hyperparameter settings for experiments.

Synthetic-to-Real Generalization
rank		lr		Params.	Trained on GTAV
main	sub	main	sub	Params.	$\rightarrow$ Citys.	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
32	2	1e-4	5e-5	13.4M	72.03	60.75	72.06	68.28
32	4	1e-4	5e-5	14.2M	73.01	61.77	72.07	68.95
32	8	1e-4	5e-5	15.7M	72.83	61.13	70.81	68.26
32	16	1e-4	5e-5	18.9M	71.20	61.16	70.87	67.74

Table 9: Domain generalization results (mIoU %) for RecycleLoRA with varying rank configurations for its Main and Sub Adapters, under the synthetic-to-real setting (G

\rightarrow

{C, B, M}). Bold and underlined indicate best and second-best results.

Real-to-Real Generalization
rank		lr		Params.	Trained on Cityscapes
main	sub	main	sub	Params.	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
32	2	1e-4	5e-5	13.4M	66.65	78.14	72.10
32	4	1e-4	5e-5	14.2M	66.76	76.64	71.70

Table 10: Domain generalization results (mIoU %) for RecycleLoRA with varying rank configurations for its Main and Sub Adapters, under the real-to-real setting (C

\rightarrow

{B, M}). Bold and underlined indicate best and second-best results.

Synthetic-to-Real Generalization
rank		lr		Params.	Trained on GTAV
main	sub	main	sub	Params.	$\rightarrow$ Citys.	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
32	4	1e-4	1e-4	14.2M	72.23	60.50	70.60	67.78
32	4	1e-4	5e-5	14.2M	73.01	61.77	72.07	68.95
32	4	1e-4	1e-5	14.2M	72.55	61.10	70.55	68.07

Table 11: Domain generalization results (mIoU %) for RecycleLoRA with varying learning rate configurations for its Main and Sub Adapters, under the synthetic-to-real setting (G

\rightarrow

{C, B, M}). Bold and underlined indicate best and second-best results.

Synthetic-to-Real Generalization
Initialization	Backbone	Trained on GTAV
Initialization	Backbone	$\rightarrow$ Citys.	$\rightarrow$ BDD	$\rightarrow$ Map.	Avg.
Kaiming unif.	DINOv2-L	68.96	60.57	69.41	66.31
SVD	DINOv2-L	70.40	60.35	70.93	67.23
RRQR	DINOv2-L	73.01	61.77	72.07	68.95

Table 12: Domain generalization results (mIoU %) for the dual-adapter framework with different initialization strategies, under the synthetic-to-real setting (G

\rightarrow

{C, B, M}). Bold and underlined indicate best and second-best results.

Figure 6: Qualitative comparison of semantic segmentation on the Cityscapes. All models were trained on the GTAV.

Figure 7: Qualitative comparison of semantic segmentation on the Mapillary. All models were trained on the GTAV.

Domain Generalization for Semantic Segmentation

See all 14 papers

LASA: Language-and-Source-Anchored Alignment for Domain Generalized Semantic Segmentation
Aug 9, 2026Jinhong Zhu, Weiqi Yan, Shengchuan Zhang +1Domain Generalization Domain Generalization for Semantic Segmentation
Xiamen University · Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China · Xiamen, Fujian, China
LoCA: Spatially-Aware Low-Rank Convolutional Adaptation of Vision Foundation Models
Jul 8, 2026Sojung An, Junha Lee, Sujeong You +2Vision Foundation Model Adaptation Low-Rank Adaptation
Korea University, Seoul, Republic of Korea · Korea Institute of Industrial Technology, Ansan, Republic of Korea · Seoul National University, Seoul, Republic of Korea
Sparse All-Layer Connector for Domain Generalised Semantic Segmentation
Oct 11, 2026Ruoyu Guo, Xin Kun Lin, Maurice Pagnucco +1Domain Generalization for Semantic Segmentation Domain Generalization
School of Computer Science and Engineering, University of New South Wales

Aug 9, 2026J/K move · Enter open · S save

LASA: Language-and-Source-Anchored Alignment for Domain Generalized Semantic Segmentation

Jinhong Zhu, Weiqi Yan, Shengchuan Zhang, Liujuan Cao

Xiamen University · Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China · Xiamen, Fujian, China

Domain Generalization Domain Generalization for Semantic Segmentation Semantic Segmentation

Open paper AskPDF arXiv

Domain Generalization Semantic Segmentation (DGSS) focuses on generalizing knowledge from labeled source domains to unseen target domains where data is unavailable during the training phase. While conventional methods utilize style randomization or feature normalization to mitigate domain shifts, they often impair feature integrity. Specifically, style randomization distorts the underlying feature manifold due to its coarse-grained nature, while feature normalization suppresses discriminative, domain-sensitive semantic details owing to its rigid design. To address these limitations, we propose the Language-and-Source-Anchored Alignment (LASA) framework, which comprises three synergistic components: Text-and-Source-Guided Style Transfer (TSGST), Domain-Aware Query Adapter (DAQA), and Domain-Aware Decoder Optimizer (DADO). Concretely, the TSGST module addresses manifold distortion by utilizing source features as structural anchors and vision-language model (VLM) priors as fine-grained guidance. To restore suppressed discriminative and domain-sensitive details, the DAQA module recalibrates object queries via categorical guidance and domain-aware signatures, while the DADO module aligns the resulting query distributions with a shared classifier to ensure consistent categorical responses across domains. Extensive experiments on challenging benchmarks demonstrate that our method significantly outperforms state-of-the-art approaches.

Abstract

Domain Generalization for Semantic Segmentation

Explore similar work

LASA: Language-and-Source-Anchored Alignment for Domain Generalized Semantic Segmentation

LoCA: Spatially-Aware Low-Rank Convolutional Adaptation of Vision Foundation Models

Sparse All-Layer Connector for Domain Generalised Semantic Segmentation