LiDAR-based perception is critical for autonomous driving due to its robustness to poor lighting and visibility conditions. Yet, current models operate under the closed-set assumption and often fail to recognize unexpected out-of-distribution (OOD) objects in the open world. Existing OOD scoring functions exhibit limited performance because they ignore the pronounced class imbalance inherent in LiDAR OOD detection and assume a uniform class distribution. To address this limitation, we propose the Neural Distribution Prior (NDP), a framework that models the distributional structure of network predictions and adaptively reweights OOD scores based on alignment with a learned distribution prior. NDP dynamically captures the logit distribution patterns of training data and corrects class-dependent confidence bias through an attention-based module. We further introduce a Perlin noise-based OOD synthesis strategy that generates diverse auxiliary OOD samples from input scans, enabling robust OOD training without external datasets. Extensive experiments on the SemanticKITTI and STU benchmarks demonstrate that NDP substantially improves OOD detection performance, achieving a point-level AP of 61.31% on the STU test set, which is more than 10× higher than the previous best result. Our framework is compatible with various existing OOD scoring formulations, providing an effective solution for open-world LiDAR perception.
Figures & tables
Figure 1: OOD objects are hazardous for LiDAR perception models because they are often misclassified as known categories. For example, an armchair on the road is incorrectly predicted as “road” by a closed-set model. The proposed NDP correctly assigns high OOD scores (red) to the road hazard (armchair) and to the roadside anomalies (rubbish bins).
Figure 2: Overview of the proposed Neural Distribution Prior (NDP) framework. Given an input point cloud, synthetic OOD samples are generated using the Perlin Raise procedure and jointly trained with in-distribution data. A sparse UNet [ 17 ] extracts point features, which are processed by an MLP to produce class logits for OOD detection, while the transformer decoder utilizes multi-level features from the UNet to generate mask-based predictions for in-distribution segmentation. The NDP module projects the input logits into a latent embedding space and performs cross-attention with a learnable prior matrix ψ , producing a reweighting term W(fΘ,ψ) that adjusts the static OOD score to generate calibrated uncertainty maps.
Method
Auxiliary OOD Data
Point-Level OOD
Object-Level OOD
AUROC ↑
FPR@95 ↓
AP ↑
RecallQ ↑
SQ ↑
RQ ↑
UQ ↑
PQ ↑
Deep Ensemble [ 33 ]
✗
90.93
37.34
6.94
17.70
79.96
9.10
14.15
7.27
MC Dropout [ 56 ]
✗
65.76
79.82
0.17
3.54
74.36
3.48
2.63
2.59
MaxLogit [ 26 ]
✗
87.27
68.76
2.02
26.64
79.26
2.06
21.12
1.63
Void Classifier [ 4 ]
✓
89.77
79.50
2.62
17.35
81.27
8.98
14.10
7.30
RbA [ 48 ]
✗
73.00
100.0
1.64
21.84
78.58
2.75
17.16
2.16
Table 1: Anomaly segmentation performance on the validation set of the STU benchmark. All methods use the Mask4Former [ 68 ] architecture. NDP substantially improves both point-level and object-level OOD detection performance compared with existing methods.
Method
Auxiliary OOD Data
Point-Level OOD
Object-Level OOD
AUROC ↑
FPR@95 ↓
AP ↑
RecallQ ↑
SQ ↑
RQ ↑
UQ ↑
PQ ↑
Deep Ensemble [ 33 ]
✗
86.74
58.05
5.17
16.75
84.49
10.43
14.16
8.81
MC Dropout [ 56 ]
✗
61.51
82.37
0.11
2.25
86.72
1.95
2.14
1.86
MaxLogit [ 26 ]
✗
84.53
81.49
0.95
26.14
83.06
2.13
21.71
1.77
Void Classifier [ 4 ]
✓
85.99
78.60
3.92
17.64
84.40
8.19
14.89
6.91
RbA [ 48 ]
✗
66.38
100.0
0.81
24.04
83.28
3.23
20.02
2.69
Table 2: Anomaly segmentation performance of the test set of the STU benchmark. All methods use the Mask4Former [ 68 ] architecture. NDP consistently achieves state-of-the-art performance under both point-level and object-level evaluations.
Figure 3: Visualization of OOD score map on the STU benchmark. Points are labeled as inlier , anomaly , and unlabeled . A continuous color bar indicates the predicted likelihood of each point being ID or OOD . Our approach yields precise and coherent anomaly masks while maintaining a low false-positive rate on inlier regions.
Method
Aux. Data
AUROC ↑
FPR@95 ↓
AP ↑
MaxLogit [ 26 ]
✗
90.73
47.63
53.86
RbA [ 48 ]
✗
78.86
100.00
55.23
UEM [ 49 ]
✓
93.15
37.07
61.73
REAL* [ 7 ]
✓
84.90
-
20.08
APF* [ 35 ]
✗
85.60
-
36.10
LiON* [ 66 ]
✓
92.69
-
44.68
Table 3: Outlier detection performance on semanticKITTI outlier classes. Object-level performance is not applicable because outlier classes do not have instance masks. * denotes model using Cylinder3D [ 72 ] architecture. Other methods use the Mask4Former [ 68 ] architecture.
Method
STU (PQ)
SemanticKITTI (PQ)
Mask4Former-closed-set [ 46 ]
52.73
60.72
Mask4Former-void [ 4 ]
26.96
47.97
Mask4Former-NDP
52.37
59.38
Table 4: In-distribution panoptic segmentation on the validation sets of STU [ 50 ] and SemanticKITTI [ 1 ] . Our method outperforms other approaches that involve OOD training.
Method
AUROC ↑
FPR@95 ↓
AP ↑
Void Training
94.54
22.51
2.09
Mixed Training
98.47
8.28
11.03
w/o SOE
99.32
1.92
67.10
w/o NDP
99.18
3.02
58.69
NDP-EE
99.53
1.43
74.24
Table 5: Ablation study of the NDP framework. All components contribute positively to the overall performance.
Method
AUROC ↑
FPR@95 ↓
AP ↑
w/o NDP
98.50
4.33
56.88
Statistics Prior [ 29 ]
97.99
3.72
23.87
Balanced Energy [ 16 ]
98.47
3.17
57.05
NDP-Energy
99.37
2.35
66.54
Table 6: Comparison with the statistic class distribution prior using energy-based OOD score. NDP largely outperforms the statistics class distribution prior and static OOD score.
β in SOE
AUROC ↑
FPR@95 ↓
AP ↑
0.7
99.26
1.53
68.00
0.8
99.36
2.35
72.73
0.9
99.53
1.43
74.24
1.0
99.36
1.61
73.06
Table 7: Ablation study of β in soft outlier exposure. SOE is not sensitive to the choice of hyperparameter.
Figure 4: Additional visualization of OOD score map on the STU benchmark with image reference. Points are labeled as inlier , anomaly , and unlabeled . Our approach yields precise and coherent anomaly masks while maintaining a low false-positive rate on inlier regions.
Figure 5: Additional visualization of OOD score map on the STU benchmark. Points are labeled as inlier , anomaly , and unlabeled . Our approach yields precise and coherent anomaly masks while maintaining a low false-positive rate on inlier regions.
Figure 6: Visualization of OOD score map on the SemanticKITTI. Points are labeled as inlier , anomaly , and unlabeled . Our approach yields precise and coherent anomaly masks while maintaining a low false-positive rate on inlier regions.
d
AUROC ↑
FPR@95 ↓
AP ↑
8
99.42
1.21
70.29
16
99.53
1.43
74.24
32
99.20
1.67
70.14
Table 8: Ablation study of template size d in NDP matrix ψ , where d is the dimensionality of the vectors stored in ψ as the learnable prior.
Method
AUROC ↑
FPR@95 ↓
AP ↑
Static Extended Energy
98.33
2.94
58.36
NDP-EE
99.19
2.89
70.29
Table 9: OOD Detection Results of NDP-EE using various backbones
Method
void
car
truck
bicycle
person
road
sidewalk
parking
building
vegetation
trunk
terrain
fence
pole
traffic sign
PQ
Mask4Former [ 68 ]
–
80.99
37.28
47.65
80.99
71.46
17.74
0.0
84.08
89.73
29.34
30.79
47.6
59.62
60.96
52.73
Mask4Former-void [ 4 ]
0.07
23.88
20.78
1.01
43.30
38.24
20.03
11.11
48.45
43.09
20.20
17.31
30.80
27.26
33.16
26.96
Mask4Former-NDP
–
77.42
48.58
51.47
76.05
40.37
12.37
0.0
90.83
92.84
31.00
65.09
36.14
51.28
59.77
52.37
Table 10: In-distribution per-class performance of the methods on the validation set of STU [ 50 ] dataset. Our model retains comparable panoptic segmentation performance to standard Mask4Former training.
Method
void
car
truck
bicycle
motorcycle
other vehicle
person
bicyclist
motorcyclist
road
sidewalk
parking
other ground
building
vegetation
trunk
terrain
fence
pole
traffic sign
PQ
Mask4Former [ 68 ]
–
93.53
59.39
62.55
64.82
54.36
79.61
89.16
25.01
93.24
77.90
28.79
0.0
87.27
87.28
51.08
59.92
24.85
56.76
58.14
60.72
Mask4Former-void [ 4 ]
6.08
74.36
47.00
32.19
43.34
33.30
42.90
68.75
00.33
93.35
77.07
19.01
0.0
82.77
81.34
47.56
56.94
19.98
54.48
36.82
47.97
Mask4Former-NDP
–
93.22
60.77
60.17
68.99
56.62
80.25
87.93
0.0
92.90
77.29
25.44
0.0
87.28
86.88
52.01
59.89
23.61
58.31
56.73
59.38
Table 11: In-distribution per-class performance of the methods on validation sets of SemanticKITTI [ 1 ] . Our model retains comparable panoptic segmentation performance to standard Mask4Former training.
Figure 7: Class distribution in the SemanticKITTI dataset. In the dataset, vegetation, road, and sidewalk account for most points. Vegetation alone contributes more than a quarter of all points, and the top four to five classes collectively exceed half of the dataset. In contrast, many classes such as motorcyclist, bicyclist, bicycle, person, and traffic sign appear in extremely small proportions. These categories often fall below one percent of the total point count.
Figure 8: Class distribution in the STU dataset. Using the sequence 201 with full annotation as an example, the scene contains over 90 million LiDAR points in total. A few dominant classes, such as vegetation, road, and sidewalk, occupy most of the point cloud. Rare classes like person, traffic-sign, and bicycle comprise less than 0.5% of all points.
Figure 9: Range-view visualization of Perlin Raise–generated OOD samples. Blue regions denote synthetic anomalies.
LiDAR-based 3D object detection plays a critical role for reliable and safe autonomous driving systems. However, existing detectors often produce overly confident predictions for objects not belonging to known categories, posing significant safety risks. This is caused by so-called out-of-distribution (OOD) objects, which were not part of the training data, resulting in incorrect predictions. To address this challenge, we propose ALOOD (Aligned LiDAR representations for Out-Of-Distribution Detection), a novel approach that incorporates language representations from a vision-language model (VLM). By aligning the object features from the object detector to the feature space of the VLM, we can treat the detection of OOD objects as a zero-shot classification task. We demonstrate competitive performance on the nuScenes OOD benchmark, establishing a novel approach to OOD object detection in LiDAR using language representations. The source code is available at https://github.com/uulm-mrm/mmood3d.
Michael Kösel, Marcel Schreiber, Michael Ulrich +2
Institute of Measurement, Control, and Microtechnology, Ulm University, Germany · Robert Bosch GmbH, Corporate Research, 71272 Renningen, Germany
Understanding the surrounding environment is fundamental in autonomous driving and robotic perception. Distinguishing between known classes and previously unseen objects is crucial in real-world environments, as done in Anomaly Segmentation. However, research in the 3D field remains limited, with most existing approaches applying post-processing techniques from 2D vision. To cover this lack, we propose a new efficient approach that directly operates in the feature space, modeling the feature distribution of inlier classes to constrain anomalous samples. Moreover, the only publicly available 3D LiDAR anomaly segmentation dataset contains simple scenarios, with few anomaly instances, and exhibits a severe domain gap due to its sensor resolution. To bridge this gap, we introduce a set of mixed real-synthetic datasets for 3D LiDAR anomaly segmentation, built upon established semantic segmentation benchmarks, with multiple out-of-distribution objects and diverse, complex environments. Extensive experiments demonstrate that our approach achieves state-of-the-art and competitive results on the existing real-world dataset and the newly introduced mixed datasets, respectively, validating the effectiveness of our method and the utility of the proposed datasets. Code and datasets are available at https://simom0.github.io/lido-page/.
3D semantic occupancy prediction is central to autonomous driving, yet current methods are vulnerable to long-tailed class bias and out-of-distribution (OOD) inputs, often overconfidently assigning anomalies to rare classes. We present ProOOD, a lightweight, plug-and-play method that couples prototype-guided refinement with training-free OOD scoring. ProOOD comprises (i) prototype-guided semantic imputation that fills occluded regions with class-consistent features, (ii) prototype-guided tail mining that strengthens rare-class representations to curb OOD absorption, and (iii) EchoOOD, which fuses local logit coherence with local and global prototype matching to produce reliable voxel-level OOD scores. Extensive experiments on five datasets demonstrate that ProOOD achieves state-of-the-art performance on both in-distribution 3D occupancy prediction and OOD detection. On SemanticKITTI, it surpasses baselines by +3.57% mIoU overall and +24.80% tail-class mIoU; on VAA-KITTI, it improves AuPRCr by +19.34 points, with consistent gains across benchmarks. These improvements yield more calibrated occupancy estimates and more reliable OOD detection in safety-critical urban driving. The source code is publicly available at https://github.com/7uHeng/ProOOD.
Yuheng Zhang, Mengfei Duan, Kunyu Peng +5
Hunan University · Karlsruhe Institute of Technology · INSAIT, Sofia University “St. Kliment Ohridski”