UniLoop: Unified Loop Closure Detection from Retrieval to Geometric Verification and Relative Pose Estimation
Authors: Jingwen Yu, Pengyu Wang, Jin Wu, Weinan Chen, Ping Tan, Hong Zhang
Organizations: CKS-Robotics Institute, Hong Kong University of Science and Technology, Hong Kong, China. · Department of Electronic and Electrical Engineering, Southern University of Science and Technology, Shenzhen, China. · School of Intelligent Science and Technology, University of Science and Technology Beijing, Beijing, China. · Biomimetics and Intelligent Robotics Lab (BIRL), School of Electromechanical Engineering, Guangdong University of Technology, Guangzhou, China.
Loop closure detection conventionally cascades separate models for global retrieval, geometric verification, and relative pose estimation, leading to redundant computation and disconnected visual representations. We propose UniLoop, a unified framework that reuses the hierarchical representations of the pretrained geometric foundation model throughout this pipeline. A shared image encoder produces features for compact global descriptors; the same features are passed to the cross-view decoder for pairwise verification; and the original geometric predictions provide relative camera constraints for pose-graph optimization. Experiments on place recognition, loop verification, and relative pose estimation show that UniLoop improves loop-closure robustness and accuracy while reducing redundant backbone computation relative to modular pipelines. UniLoop achieves the best Recall@10 on MSLS (96.6) and Pitts30k (99.3) among the compared methods, while its verification branch reaches 95.63 AP and 96.29 ROC-AUC, surpassing the existing geometric verification baselines.
Figures & tables
Fig. 2: Architecture of UniLoop for a candidate image pair ( Ii , Ij ). A shared-weight MASt3R ViT encoder processes each image once. The SALAD head aggregates the resulting single-view tokens into normalized global descriptors gi and gj for candidate retrieval. For each shortlisted pair, the same tokens are passed to the cross-view decoder to obtain pair-conditioned features. A lightweight verification head predicts the loop probability pijloop , while the retained MASt3R’s camera and pointmap heads recover the geometry used to construct relative-pose constraints for pose-graph optimization. Blue and green denote the two input views and their corresponding feature streams.
Trainable Blks
Distill.
Resolution
Avg. Rel. Imp. (%)
MSLS
Pitts30k-val
Pitts30k-test
R@1
R@5
R@10
R@1
R@5
R@10
R@1
R@5
R@10
0
✗
224/224
–
76.35
84.59
87.97
90.17
97.24
98.26
88.13
94.06
95.58
2
✗
224/224
3.09
83.65
89.86
92.55
92.55
98.05
98.82
89.60
94.63
96.02
2
✓
224/224
3.75
83.65
91.22
93.65
93.48
98.38
99.07
90.18
95.03
96.45
2
✗
224/512
4.29
87.16
93.24
93.92
92.18
97.66
98.62
90.17
95.30
96.33
4
✗
224/224
3.50
84.32
90.95
92.84
93.20
97.95
98.70
90.05
94.76
96.07
TABLE II: Retrieval branch ablation on MSLS and Pitts30k-val, and Pitts30k-test. “Blks” is the number of fine-tuned MASt3R encoder blocks; “Dist.” indicates whether distillation is used (✓/✗). “Avg. Rel. Imp.” is the mean relative improvement across the nine reported metrics with respect to the first row.
Method
AP
ROC-AUC
MR
IR
SALAD
94.85
94.97
0.25
Ours (VPR-branch)
92.41
92.88
0.31
GV
SP+LG
84.67
89.87
0.0
eLoFTR
86.94
91.31
0.0
NC
Train from scratch
93.22
94.36
0.12
\drsh\ + Hard Negatives
94.00
95.49
0.45
TABLE III: Comparison and ablation of the loop verification branch. IR denotes image retrieval, GV denotes geometric verification, and NC denotes the proposed neural classifier.
Feedforward foundation models have recently shown remarkable 3D reconstruction capabilities. However, existing models exhibit large tracking drift in long-context streaming reconstruction due to error accumulation. In this paper, we revisit loop closure with streaming reconstruction foundation models to enable accurate, drift-free, kilometer-scale reconstruction. Specifically, our method detects loop candidates through global descriptor retrieval, and constructs loop-conditioned windows to estimate the relative poses between looped frames. Given the observation that our adopted streaming reconstruction backbone produces a globally consistent scale, we optimize all frame poses on the SE(3) manifold with sequential and loop closure constraints, avoiding the pose graph optimization on the Sim(3) or higher-dimensional SL(4) manifolds employed in prior works. Extensive experiments show that our method reduces drift and produces consistent geometry on kilometer-scale sequences, significantly outperforming the state of the art. Code is available at https://github.com/MoyangLi00/CLoSeR.git.
Loop closure is essential to reduce drift and build globally consistent maps in large-scale environments. However, reliable loop closure with only geometric information from, e.g., a LiDAR sensor, remains challenging due to the difficulty of constructing discriminative geometric features. We present PinNet, a neural network that produces local geometric descriptors from point clouds for place recognition and scanto-scan registration. PinNet incorporates a neural network that generates keypoints and their corresponding descriptors, together with a plane-based geometric self-attention module that models inter-keypoint spatial relationships to enhance descriptor discriminability for loop-closure detection and point-cloud registration. The approach is comprehensively evaluated on multiple datasets collected with different LiDAR sensors. Experimental results demonstrate strong place-recognition performance, precise relative pose estimation, and successful single-shot localization in different environments.
Yanlong Ma, Nakul S. Joshi, Christa S. Robison +2
University of California, Los Angeles, Los Angeles, CA, USA. · DEVCOM Army Research Laboratory (ARL), Adelphi, MD, USA.
Visual place recognition (VPR) is a key enabler of accurate localization and long-term autonomous navigation in robotics applications, such as loop closure detection for simultaneous localisation and mapping (SLAM). However, real-world VPR deployment relies on selecting an image matching threshold that balances precision and recall. These thresholds are typically tuned using labeled validation data and fixed during deployment, making them unreliable under environmental changes where ground truth is unavailable. This is particularly problematic in safety-critical robotics, where accepting a false loop closure can corrupt the estimated trajectory and map. In this work, we introduce Visual Place Recognition Auditing, an independent post-retrieval verification framework that leverages Vision-Language Models (VLMs) to assess retrieved matches by reasoning jointly over query and candidate images. Unlike conventional verification methods, our approach performs instance-level verification without requiring architecture-specific confidence measures, dataset-dependent thresholds, or prior knowledge of the deployment environment. We evaluate our method on six benchmark datasets using five state-of-the-art VPR methods and four VLMs. Results show that VLM-based auditing improves recall@1 by 13.6% on average as compared to state-of-the-art methods while reducing false acceptance rates to 12%, maintaining precision above 95% and coverage above 75%.
Sania Waheed, Michael Milford, Sarvapali D. Ramchurn +1
School of Electronics and Computer Science, University of Southampton, Southampton SO17 1BJ, U.K. · School of Electrical Engineering and Computer Science, Queensland University of Technology, Brisbane, QLD 4000, Australia · School of Computer Science and Electronic Engineering, University of Essex, CO4 3SQ Colchester, U.K.