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 ℓ2 norm
Mean ℓ2 norm
Average
Maximum
(selected)
(non‑selected)
ratio
ratio
All layers
0.138245
0.113566
1.22 ×
1.63 ×
Table 1: Comparison of ℓ2 -norm statistics between selected and non-selected columns in LoRA matrix A after training.
Effective Rank & Rank‑Efficiency Statistics
Target rank
RecycleLoRA (Ours)
SoMA
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 A . Right: pairwise similarity among columns of B . Darker blue colors represent lower similarity.
Figure 4: Block-wise subspace similarity ϕ 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
→ Citys.
→ BDD
→ 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
→ BDD
→ 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.
→ Citys.
→ BDD
→ 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 ( ↑ 0.09)
61.77 ( ↑ 0.55)
72.07 ( ↑ 0.32)
68.95 ( ↑ 0.32)
Table 5: Ablation study on the Main and Sub Adapter components of RecycleLoRA. Best results in bold .
Method
lr.
→ Citys.
→ BDD
→ Map.
Avg.
Main Adapter
1e-4
72.92
61.22
71.75
68.63
Main Adapter
5e-5
69.46 ( ↓ 3.46)
62.22 ( ↑ 1.00)
68.23 ( ↓ 3.52)
66.64 ( ↓ 1.99)
LoRA
1e-4
70.36
60.21
69.95
66.84
LoRA
5e-5
69.40 ( ↓ 0.96)
59.62 ( ↓ 0.59)
69.20 ( ↓ 0.75)
66.07 ( ↓ 0.77)
Sub Adapter
1e-4
68.60
60.74
67.54
65.63
Sub Adapter
5e-5
70.64 ( ↑ 2.04)
60.56 ( ↓ 0.18)
71.11 ( ↑ 3.57)
67.44 ( ↑ 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
→ Citys.
→ BDD
→ 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
→ Citys.
→ BDD
→ 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 → {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
→ BDD
→ 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 → {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
→ Citys.
→ BDD
→ 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 → {C, B, M}). Bold and underlined indicate best and second-best results.
Synthetic-to-Real Generalization
Initialization
Backbone
Trained on GTAV
→ Citys.
→ BDD
→ 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 → {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 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.
Jinhong Zhu, Weiqi Yan, Shengchuan Zhang +1
Xiamen University · Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China · Xiamen, Fujian, China
Pre-trained Vision Foundation Models (VFMs) provide strong visual representations for diverse downstream tasks. The key challenge of VFM adaptation stems from the prohibitive costs of full fine-tuning and catastrophic forgetting. To address this, Low-Rank Adaptation (LoRA) has emerged as the prevailing paradigm for Parameter-Efficient Fine-Tuning (PEFT). However, LoRA is typically designed for transformer self-attention layers parameterized by 2D matrices. Since convolutional kernels inherently couple spatial and channel information within a 4D tensor, forcing them into a monolithic 2D matrix disrupts the inherent spatial topology. In this paper, we propose Low-Rank Convolutional Adaptation (LoCA), a convolution-aware PEFT framework that addresses spatial-channel entanglement by decoupling channel and spatial adaptation. LoCA introduces a low-rank channel adaptation for dense cross-channel mixing and refines spatial bases extracted from pre-trained kernels via Singular Value Decomposition (SVD). Experimental results show that LoCA preserves pre-trained spatial priors and achieves competitive or state-of-the-art performance across fine-grained classification, domain-generalized semantic segmentation, and generative benchmarks.
Sojung An, Junha Lee, Sujeong You +2
Korea University, Seoul, Republic of Korea · Korea Institute of Industrial Technology, Ansan, Republic of Korea · Seoul National University, Seoul, Republic of Korea
Domain generalised semantic segmentation (DGSS) has recently benefited from combining multiple foundation models, such as vision-language and vision foundation models, to leverage their diverse representations for generalisation. Existing multi-model approaches share a common design choice: cross-model fusion is restricted to depth-aligned layers, implicitly assuming that mutually beneficial information resides at the same depth across models. However, foundation models pretrained under different objectives may develop distinct representation hierarchies, making the optimality of depth-aligned fusion questionable. Moreover, we observe that the foundation models exhibit different depth-wise domain sensitivity. Motivated by these observations, we propose a Sparse All-Layer Connector (SALC) that enables each layer to access information from preceding layers of another model. As the number of accessible layers grows with depth, dense aggregation may hurt generalisation. SALC therefore learns to adaptively select a sparse subset of informative layers. We further introduce a candidate dropout regularisation that strengthens sparsity and encourages SALC to explore diverse layers, leading to more robust selection. Across four foundation model combinations and three DGSS evaluation settings, the proposed design demonstrates stronger generalisation than conventional depth-aligned fusion. The code is available at https://github.com/RuoyuGuo/SALC.
Ruoyu Guo, Xin Kun Lin, Maurice Pagnucco +1
School of Computer Science and Engineering, University of New South Wales