Pith. sign in

REVIEW 5 major objections 7 minor 1 cited by

OoDDINO:A Multi-level Framework for Anomaly Segmentation on Complex Road Scenes

T0 review · 5 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read OoDDINO claims that bolting an object-level anomaly detector onto frozen pixel-wise segmenters, then cutting the score map with two region-specific thresholds instead of one global threshold, sharply reduces false positives and raises AP…

desk verdict Architecture is coherent and the plug-in idea is useful, but test-set-tuned hyperparameters and contradictory tables make the reported gains unverifiable. read the letter →

arxiv 2507.01455 v2 pith:4OWASDEO submitted 2025-07-02 cs.CV

Yuxing Liu , Ji Zhang , Zhou Xuchuan , Jingzhong Xiao , Huimin Yang , Jiaxin Zhong This is my paper · ORCID
classification cs.CV MSC 68T4568U10
keywords anomalysegmentationopen-setobjectdetectionadaptivedual-thresholdnetworkorthogonaluncertainty-awarefusionout-of-distributionroadsceneunderstandingautonomousdriving
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that the two chronic failures of pixel-wise anomaly segmentation — fragmented masks and a one-size-fits-all global score threshold — can be repaired without retraining the pixel-level model. Its recipe is a two-stage cascade: an open-set object detector (GroundingDINO, enhanced by fusing three uncertainty cues under an orthogonality penalty) proposes anomalous regions, and a learned Adaptive Dual-Threshold Network cuts the frozen pixel-wise score map once inside those proposals and once outside them. On the SMIYC and RoadAnomaly benchmarks the cascade lowers false-positive rates by roughly a third to an order of magnitude across baselines (for example RPL's FPR95 from 11.7% to 7.8% on AnomalyTrack and from 0.6% to 0.06% on ObstacleTrack) while raising AP and component-level F1. If correct, the framework is a plug-in upgrade: existing anomaly segmentation models can gain object-level spatial priors by adding the detection stage and the dual thresholds rather than redesigning their scoring. The whole gain, however, is hostage to the detector's proposals — a fact the paper's own ablation illustrates, since bolting on raw GroundingDINO without the two new modules makes every metric worse.

What carries the argument

The key machinery is the two-stage division of a single thresholding problem into an object-proposal problem and a regional-calibration problem. On the detection side, OUAFS (Orthogonal Uncertainty-Aware Fusion Strategy) serially fuses three uncertainty maps — a semantic segmentation map, a softmax entropy map, and a softmax distance map — with the visual features of GroundingDINO, ordering the fusions by measured contribution and applying an orthogonality loss that penalises redundancy between the fused features. On the segmentation side, ADT-Net (Adaptive Dual-Threshold Network) takes the detector's foreground and background masks and the frozen pixel-wise anomaly score map, normalises scores piecewise against the mean score of each region, and learns two thresholds from architecturally identical predictors, one for foreground and one for background. A cumulative-distribution-function relaxation makes the hard thresholding operation differentiable during training, while a cross-entropy plus threshold-divergence loss keeps the two cuts apart. The mechanism works because it converts an arbitrarily distributed score map into a form where two learned cuts can adapt to each image.

What would settle it

Filter a test set of road scenes to the images where the OUAFS-fused detector outputs no bounding box overlapping a ground-truth anomaly, and measure the cascade's FPR95 and AP on that subset alone. If the cascade is not at least as good as the frozen baseline there, the claim that the framework can boost frozen segmenters is falsified; the paper's own RPL+GD ablation row shows such images exist.

Watch

Extended reading notes

Core claim

The paper's central claim, stated on its own terms: anomaly segmentation should be solved coarse-to-fine, with an object-level detector deciding where anomalies are and a pixel-level scorer deciding how much of each detected region is anomalous. The authors claim to be the first to apply an open-set detection model to anomaly segmentation, and their evidence is that the two new modules convert a frozen pixel-wise scorer into a substantially better segmenter. Specifically, OUAFS raises the detector's ability to localize anomalies of all sizes (mAP improves with each added uncertainty cue and with the sequential ordering), and ADT-Net's region-adaptive thresholds, trained with a cumulative-distribution-function relaxation of hard binarization, are what cut FPR95 while preserving recall. The paper reports that across thirteen metrics on three datasets, the framework improved twelve for RPL and eleven for RbA, with the best configuration reaching 98.9% AuROC on RoadAnomaly.

Load-bearing premise

The framework stands or falls on the detector's proposals: if the OUAFS-fused GroundingDINO fails to place a box over an actual anomaly, or places spurious boxes over normal road content, ADT-Net's region-specific thresholds inherit that mistake and the whole cascade can end up worse than the frozen baseline — the paper's own ablation shows raw GroundingDINO integration (without ADT-Net or OUAFS) raising RPL's FPR95 on RoadAnomaly from 17.74% to 28.50%.

Editorial extensions

If this is right

  • Frozen pixel-wise anomaly segmentation models can be upgraded as a plug-in: OoDDINO improves AP and lowers FPR95 for RPL and RbA, and the appendix reports the same transfer to PEBAL and Mask2Anomaly (Mask2Anomaly FPR95 on RoadAnomaly drops from 13.45% to 7.33%).
  • False-positive control is the main, consistent win: FPR95 falls on every dataset and baseline, and in the small-obstacle and night-time conditions of ObstacleTrack it falls by an order of magnitude (RPL from 0.6% to 0.06%).
  • Component-level mask quality improves alongside pixel metrics, because the detector's boxes restore object coherence: PPV and F1 rise substantially (RPL F1 from 30.1 to 56.1 on AnomalyTrack and from 56.6 to 86.5 on ObstacleTrack).
  • The gain decomposes cleanly: ADT-Net alone repairs most of the damage caused by raw detector errors (AnomalyTrack FPR95 from 21.25% down to 6.50%), while OUAFS lifts detector mAP across small, medium, and large anomalies.
  • The upgrade carries a real compute cost that the paper discloses: the full model uses about 660M parameters and 410.88 GFLOPs versus 168M and 32.1 GFLOPs for RPL, with inference speed dropping from 4.56 to 3.51 FPS.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because ADT-Net only re-thresholds regions the detector proposes, the framework's gain should track detector recall: on any image where the OUAFS-fused GroundingDINO outputs no box covering the anomaly, the cascade should collapse toward (or below) the frozen baseline — a testable prediction that the paper's RPL+GD ablation row already gestures at.
  • The dual-threshold stage is essentially a learned, region-conditional calibration of whatever score map it is given, so the plug-in claim is testable beyond RPL and RbA: the same ADT-Net should improve energy-based or uncertainty-based score maps, provided the proposals cover the anomalies.
  • Since ADT-Net learns from box proposals and score statistics rather than anomaly annotations, a natural extension is unsupervised region thresholding — training the two predictors on detector pseudo-labels alone would test whether the pixel-wise scorer is needed at all during training.
  • One internal inconsistency a careful reader will notice: the RPL baseline on AnomalyTrack is reported as 11.7% FPR95 and 83.4% AP in Table 1 but as 7.18% and 88.55% in Table 3, so the exact magnitude of the claimed gains depends on which baseline row is taken as reference; the direction of the gains is consistent, but the size needs a check.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 7 minor

Summary. The paper proposes OoDDINO, a two-stage framework for anomaly segmentation in road scenes. In the first stage, an Orthogonal Uncertainty-Aware Fusion Strategy (OUAFS) integrates semantic-segmentation, softmax-entropy, and softmax-distance maps into a GroundingDINO open-set detector under an orthogonal-loss constraint. In the second stage, an Adaptive Dual-Threshold Network (ADT-Net) normalizes pixel-wise anomaly scores and learns separate foreground/background thresholds conditioned on detection proposals. The anomaly segmentation baselines RPL and RbA are kept frozen, and the paper reports substantial gains in FPR95, AP, and component-level F1 on SMIYC, RoadAnomaly, and Fishyscapes, positioning OoDDINO as a plug-in module for existing pixel-wise anomaly segmentation methods.

Significance. If the reported results are correct, the work is a useful and fairly original contribution: it is, to my knowledge, the first systematic use of an open-set detector as an object-level prior for anomaly segmentation, and the proposed two-threshold mechanism directly addresses a known weakness of global thresholding. The modular design is a genuine strength, as is the release of source code. However, the empirical evidence for the central claim currently has two load-bearing problems: the same configuration is reported with different numbers in Tables 1 and 3, and the normalization-range parameter alpha in Eq. (2) is selected on the test set in Appendix C. These issues must be resolved before the headline improvements can be trusted.

major comments (5)
  1. [Tables 1 and 3] The same RPL-based configuration is reported with inconsistent numbers on AnomalyTrack: Table 1 gives RPL an AP of 83.4 and FPR95 of 11.7, while Table 3 gives RPL an AP of 88.55 and FPR95 of 7.18; similarly, Table 1 reports RPL+Ours with AP 87.3 and FPR95 7.8, while Table 3 reports OoDDINO with AP 92.08 and FPR95 3.82. Because Tables 1 and 3 are supposed to evaluate the same framework on the same benchmark, this discrepancy changes the claimed improvement by roughly 5 AP points and 4 FPR95 points. Please reconcile the two tables and report a single consistent evaluation protocol, or explain explicitly why the numbers differ.
  2. [Appendix C, Eq. (2)] The normalization-range parameter alpha is tuned on the RoadAnomaly test set: Appendix C states that the lower bound was varied from 0.1 to 0.4 and that [0.3, 0.8] was selected because it gave the best FPR/AP. As a result, the headline numbers, including the 98.9% AuROC and the 4.78% FPR95, are the best of several test-set evaluations rather than estimates under a fixed protocol. This makes the reported gains look potentially overfit to the test benchmarks. Please fix alpha using a validation split or a pre-registered default, and report test-set performance for the single chosen value.
  3. [Sec. 3.3, Eq. (2)] The foreground mask is defined as Mfg = 1 in {1}^{H x W} and the background mask as Mbg = 1 - Mfg, which makes the background branch empty and contradicts the surrounding text, which says the masks are derived from detection proposals. As written, Eq. (2) does not specify the advertised region-adaptive mechanism. Please correct the definition to Mfg = 1_{boxes} (or an equivalent indicator over detection proposals) and define Mbg accordingly.
  4. [Sec. 4.1, Tables 1-3] All experiments are reported as point estimates without error bars, multiple seeds, or confidence intervals, on small test sets (RoadAnomaly has 60 images and AnomalyTrack has 100). Many of the claimed improvements are only a few AP or FPR95 points, so without repeated-run statistics it is unclear whether the differences are within run-to-run noise. Please add standard deviations over at least three runs, or otherwise justify the stability of the reported numbers.
  5. [Sec. 4.3, Table 3] The ablation shows that naively adding GroundingDINO to RPL degrades all metrics, e.g., RoadAnomaly FPR95 rises from 17.74 to 28.50. The final framework therefore depends critically on OUAFS and ADT-Net correcting detector errors, but the paper does not analyze cases where the detector still misses anomalies or outputs spurious boxes, nor does it report detector recall on the benchmarks. Please add a quantitative analysis of detection failures and a discussion of how the framework behaves when the object-level proposals are unreliable.
minor comments (7)
  1. [Sec. 4.1, 4.2.1] There are typos in the text: 'segemntation' in Section 4.1 and 'ObstleTrack' in Section 4.2.1 should be corrected.
  2. [References] References [9] and [10] appear to be the same paper, and [19] and [20] are also duplicates; please merge them and renumber.
  3. [Figure 3] The caption says 'the two leftmost columns display the image and its ground truth,' but the described figure and surrounding text imply more columns; please make the caption consistent with the actual figure layout.
  4. [Table 4] The Small/Medium/Large columns in Table 4 are not defined in the main text; please state what these categories refer to (e.g., object sizes or bounding-box scales) and how they are computed.
  5. [Abstract and Sec. 4] The abstract says experiments use 'two benchmark datasets,' but the paper reports results on SMIYC, RoadAnomaly, and Fishyscapes; please correct the count.
  6. [Sec. 3.2] The text does not specify which model produces the semantic segmentation map S_i used in OUAFS, nor which text prompts are given to GroundingDINO for anomaly detection; please provide these implementation details.
  7. [Sec. 2.2] The claim that this is 'the first time an open-set detection model has been applied to anomaly segmentation' is strong and should be either verified against prior work or softened.

Circularity Check

1 steps flagged · score 4.0 of 10

RoadAnomaly headline numbers are selected by test-set tuning of alpha, while other benchmarks retain independent content.

  1. fitted input called prediction [Appendix C; Eq. (2); Table 7 vs Table 2]
    "In Section 3.3, we set the parameter α in Equation 2 empirically to 0.3. Specifically, we fix the normalized score range length to 0.5 and vary the lower bound from 0.1 to 0.4 in steps of 0.1. As shown in Table 7, the score range [0.3, 0.8] yields the best results, achieving the lowest FPR of 4.78% and the highest AP of 87.13%."

    The scalar α in Eq. (2) is an ADT-Net hyperparameter. Appendix C says it was chosen by sweeping the lower bound from 0.1 to 0.4 and picking [0.3, 0.8] because it achieved the lowest FPR and highest AP on the RoadAnomaly benchmark. Those winning values, FPR 4.78 and AP 87.13, are exactly the numbers reported as RPL+Ours on RoadAnomaly in Table 2 and highlighted in Section 4.2.2. The headline RoadAnomaly result is therefore the argmax of a hyperparameter search evaluated on the same test set, not an unbiased prediction from a fixed pipeline; the fitted hyperparameter is being presented as an independent measured gain. The SMIYC and Fishyscapes results use the fixed α without a per-dataset sweep, so those benchmarks still carry independent evidence.

full rationale

The derivation chain in this paper is empirical rather than formal: frozen pixel-wise baselines are combined with a trained detector (OUAFS/GroundingDINO) and a learned thresholding network (ADT-Net). Training ADT-Net on annotated masks is ordinary supervised learning, not circular reasoning, and the framework is evaluated against external benchmarks. No load-bearing uniqueness theorem, ansatz-smuggling citation, or self-definitional equation was found. One genuine issue does fit the fitted-input-called-prediction pattern: the normalization offset α in Eq. (2) was selected by sweeping it on the RoadAnomaly benchmark, and the best sweep values are then reported as the method's RoadAnomaly performance. This makes the flagship RoadAnomaly improvement partially a product of test-set hyperparameter selection rather than an independent measurement. In addition, Table 1 and Table 3 report conflicting values for the same RPL and OoDDINO configurations on AnomalyTrack (e.g., RPL AP 83.4 vs 88.55 and FPR95 11.7 vs 7.18; OoDDINO AP 87.3 vs 92.08 and FPR95 7.8 vs 3.82), so the magnitude of the claimed gains cannot be independently verified from the paper as written; that is a correctness/reporting risk, not a circularity step. Because the core plug-in claim retains independent support from SMIYC and Fishyscapes, and no self-citation chain is load-bearing, the circularity score is moderate rather than high.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central method rests on several domain assumptions about the availability and quality of uncertainty maps and detection proposals. No new physical entities are introduced. The main free parameters are loss weights and the normalization range, with alpha selected by a reported sweep.

free parameters (6)
  • alpha = 0.3
    Lower bound of the ADT-Net normalized score interval [alpha, alpha+0.5] in Equation 2; Appendix C sweeps lower bounds 0.1 to 0.4 and selects 0.3 as best on the reported benchmark numbers.
  • delta = 0.1
    Width of the linear transition window in the CDF threshold relaxation in Equation 5, set empirically.
  • lambda_detect = 0.5
    Weight on the GroundingDINO detection loss in Equation 7; no ablation is reported for this value.
  • lambda_orth = 0.1
    Weight on the orthogonal fusion loss in Equation 7; no ablation is reported.
  • lambda_ADT = 0.1
    Weight on the ADT-Net loss in Equation 7; no ablation is reported.
  • gamma
    Coefficient of the threshold-divergence regularizer in Equation 6; its value is not reported in the paper.
assumptions (5)
  • domain assumption Anomaly score maps from frozen baselines such as RPL and RbA are meaningful pixel-level evidence for anomaly location.
    Section 4.1 keeps RPL and RbA frozen, and ADT-Net consumes their anomaly score maps as input.
  • domain assumption Object-level proposals from OUAFS-enhanced GroundingDINO reliably cover anomalous objects, so pixels inside the proposal mask can be treated as foreground for thresholding.
    Section 3.3 derives the foreground mask from detection proposals; Section 4.3 shows RPL plus GroundingDINO alone hurts performance, revealing how much the final method relies on proposal quality.
  • ad hoc to paper Semantic segmentation maps, softmax entropy maps, and softmax distance maps are available for every training and test image at each fused layer.
    Algorithm 1 requires these maps, but the paper never states which model computes them.
  • domain assumption Training on synthetic anomalous road data transfers to the real SMIYC, RoadAnomaly, and Fishyscapes benchmarks.
    Section 4.1 says the framework is trained on the synthetic dataset following S2M; no distribution-shift analysis is provided.
  • standard math The CDF-based relaxation in Equation 5 is a valid differentiable surrogate for thresholded classification.
    It is used to propagate gradients through threshold selection, and inference applies a hard 0.5 cutoff.

how reviews work

0 comments
Cite this review

Pith. "Pith review of OoDDINO:A Multi-level Framework for Anomaly Segmentation on Complex Road Scenes." pith.science (2026). https://pith.science/paper/4OWASDEO

@misc{pith2026250701455,
  author       = {Pith},
  title        = {Pith review of: OoDDINO:A Multi-level Framework for Anomaly Segmentation on Complex Road Scenes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4OWASDEO}},
  note         = {Machine review of arXiv:2507.01455}
}
read the original abstract

Anomaly segmentation aims to identify Out-of-Distribution (OoD) anomalous objects within images. Existing pixel-wise methods typically assign anomaly scores individually and employ a global thresholding strategy to segment anomalies. Despite their effectiveness, these approaches encounter significant challenges in real-world applications: (1) neglecting spatial correlations among pixels within the same object, resulting in fragmented segmentation; (2) variabil ity in anomaly score distributions across image regions, causing global thresholds to either generate false positives in background areas or miss segments of anomalous objects. In this work, we introduce OoDDINO, a novel multi-level anomaly segmentation framework designed to address these limitations through a coarse-to-fine anomaly detection strategy. OoDDINO combines an uncertainty-guided anomaly detection model with a pixel-level segmentation model within a two-stage cascade architecture. Initially, we propose an Orthogonal Uncertainty-Aware Fusion Strategy (OUAFS) that sequentially integrates multiple uncertainty metrics with visual representations, employing orthogonal constraints to strengthen the detection model's capacity for localizing anomalous regions accurately. Subsequently, we develop an Adaptive Dual-Threshold Network (ADT-Net), which dynamically generates region-specific thresholds based on object-level detection outputs and pixel-wise anomaly scores. This approach allows for distinct thresholding strategies within foreground and background areas, achieving fine-grained anomaly segmentation. The proposed framework is compatible with other pixel-wise anomaly detection models, which acts as a plug-in to boost the performance. Extensive experiments on two benchmark datasets validate our framework's superiority and compatibility over state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2507.01455 by the authors.

Figure 1
Figure 1. (a) Input image. (b)Ground Truth. (c)The anomaly [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The OoDDINO framework integrates two complementary modules: Orthogonal Uncertainty-Aware Fusion Strategy [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison of anomaly segmentation methods.The two leftmost columns display the image and its [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative results on the ObstacleTrack dataset. Left: original images. Middle: ground-truth annotations. Right: [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Qualitative results on the AnomalyTrack dataset. Left: original images. Middle: ground-truth annotations. Right: [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Segmenting Objectiveness and Task-awareness Unknown Region for Autonomous Driving

    cs.CV 2025-04 conditional novelty 6.0 of 10

    SOTA fuses SAM features with OOD confidence maps and a road mask to segment road anomalies more completely and with fewer off-road false positives than prior detectors.

Reference graph

Works this paper leans on

55 extracted references · 55 canonical work pages · cited by 1 Pith paper

  1. [1]

    Victor Besnier, Andrei Bursuc, David Picard, and Alexandre Briot. 2021. Trigger- ing Failures: Out-of-Distribution Detection by Learning from Local Adversarial Attacks in Semantic Segmentation. In Proceedings of the IEEE International Con- ference on Computer Vision

    work page 2021

  2. [2]

    Hermann Blum, Paul-Edouard Sarlin, Juan Nieto, Roland Siegwart, and Cesar Cadena. 2021. The fishyscapes benchmark: Measuring blind spots in semantic segmentation. International Journal of Computer Vision 129, 11 (2021), 3119–3135

    work page 2021

  3. [3]

    Lang, Sourabh Vora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom

    Holger Caesar, Varun Bankiti, Alex H. Lang, Sourabh Vora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. 2020. Nuscenes: A multimodal dataset for autonomous driving. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 11618–11628

    work page 2020

  4. [4]

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. 2020. End-to-end object detection with trans- formers. In Proceedings of the European Conference on Computer Vision . 213–229

    work page 2020

  5. [5]

    Robin Chan, Krzysztof Lis, Svenja Uhlemeyer, Hermann Blum, Sina Honari, Roland Siegwart, Pascal Fua, Mathieu Salzmann, and Matthias Rottmann. 2021. Segmentmeifyoucan: A benchmark for anomaly segmentation. ARXIV (2021)

    work page 2021

  6. [6]

    Robin Chan, Matthias Rottmann, and Hanno Gottschalk. 2021. Entropy maxi- mization and meta classification for out-of-distribution detection in semantic segmentation. In Proceedings of the IEEE International Conference on Computer Vision. 5128–5137

    work page 2021

  7. [7]

    Zhi-Qi Cheng, Qi Dai, Siyao Li, Teruko Mitamura, and Alexander Hauptmann

    work page

  8. [8]

    Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus En- zweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. 2016. The cityscapes dataset for semantic urban scene understanding. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    work page 2016

Show all 55 references
  1. [9]

    Giancarlo Di Biase, Hermann Blum, Roland Siegwart, and Cesar Cadena. 2021. Pixel-wise anomaly detection in complex driving scenes. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 16918–16927

    2021

  2. [10]

    Siegwart, and César Cadena

    Giancarlo Di Biase, Hermann Blum, Roland Y. Siegwart, and César Cadena. 2021. Pixel-Wise Anomaly Detection in Complex Driving Scenes. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    2021

  3. [11]

    Yarin Gal and Zoubin Ghahramani. 2016. Dropout as a Bayesian Approxima- tion: Representing Model Uncertainty in Deep Learning. In Proceedings of the International Conference on Machine Learning . 1050–1059

    2016

  4. [12]

    P. Gao, S. Geng, R. Zhang, T. Ma, R. Fang, Y. Zhang, H. Li, and Y. Qiao. 2021. Clip-adapter: Better vision-language models with feature adapters.ARXIV (2021)

    2021

  5. [13]

    Matej Grcic, Petra Bevandić, and Siniša Šegvić. 2021. Dense Anomaly Detection by Robust Learning on Synthetic Negative Data. ARXIV (2021)

    2021

  6. [14]

    Matej Grcic, Petra Bevandić, and Siniša Šegvić. 2022. DenseHybrid: Hybrid Anom- aly Detection for Dense Open-Set Recognition. In Proceedings of the European Conference on Computer Vision

    2022

  7. [15]

    Xiuye Gu, Tsung-Yi Lin, Weicheng Kuo, and Yin Cui. 2021. Open-vocabulary object detection via vision and language knowledge distillation. Learning (2021)

    2021

  8. [16]

    Jun-Yan He, Zhi-Qi Cheng, Chenyang Li, Wangmeng Xiang, Binghui Chen, Bin Luo, Yifeng Geng, and Xuansong Xie. 2023. Damo-streamnet: Optimizing stream- ing perception in autonomous driving. ARXIV (2023)

    2023

  9. [17]

    Matthias Hein, Maksym Andriushchenko, and Julian Bitterwolf. 2019. Why relu networks yield high-confidence predictions far away from the training data and how to mitigate the problem. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    2019

  10. [18]

    Dan Hendrycks and Kevin Gimpel. 2016. A baseline for detecting misclassified and out-of-distribution examples in neural networks. ARXIV (2016)

    2016

  11. [20]

    Dan Hendrycks and Kevin Gimpel. 2017. A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks. In Proceedings of the International Conference on Learning Representations

    2017

  12. [21]

    Hanzhe Hu, Jinshi Cui, and Liwei Wang. 2021. Region-aware contrastive learning for semantic segmentation. In Proceedings of the IEEE International Conference on Computer Vision. 16291–16301

    2021

  13. [22]

    K. J. Joseph, S. Khan, F. S. Khan, and V. N. Balasubramanian. 2021. Towards Open World Object Detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 5830–5840

    2021

  14. [23]

    Sanghun Jung, Jungsoo Lee, Daehoon Gwak, Sungha Choi, and Jaegul Choo

  15. [24]

    Alex Kendall and Yarin Gal. 2017. What uncertainties do we need in bayesian deep learning for computer vision? NIPS 30 (2017)

    2017

  16. [25]

    Kingma and J

    D.P. Kingma and J. Ba. 2017. Adam: A Method for Stochastic Optimization. In Proceedings of the International Conference on Learning Representations

    2017

  17. [26]

    Balaji Lakshminarayanan, Alexander Pritzel, and Charles Blundell. 2017. Sim- ple and Scalable Predictive Uncertainty Estimation using Deep Ensembles. In Proceedings of the Advances in Neural Information Processing Systems . 6402–6413

    2017

  18. [27]

    Jin-Peng Lan, Zhi-Qi Cheng, Jun-Yan He, Chenyang Li, Bin Luo, Xu Bao, Wang- meng Xiang, Yifeng Geng, and Xuansong Xie. 2023. Procontext: Exploring pro- gressive context transformer for tracking. In Proceedings of the IEEE International Conference on Acoustics, Speech, and Sign...

    2023

  19. [28]

    Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. 2018. A simple unified framework for detecting out-of-distribution samples and adversarial attacks. NIPS 31 (2018)

    2018

  20. [29]

    Chenyang Li, Zhi-Qi Cheng, Jun-Yan He, Pengyu Li, Bin Luo, Hanyuan Chen, Yifeng Geng, Jin-Peng Lan, and Xuansong Xie. 2023. Longshortnet: Exploring temporal and semantic features fusion in streaming perception. In Proceedings of the IEEE International Conference on Acoustics, ...

    2023

  21. [30]

    Xia Li, Zhisheng Zhong, Jianlong Wu, Yibo Yang, Zhouchen Lin, and Hong Liu

  22. [31]

    Chen Liang, Wenguan Wang, Jiaxu Miao, and Yi Yang. 2022. GMMSeg: Gaussian Mixture Based Generative Semantic Segmentation Models. ARXIV (2022)

    2022

  23. [32]

    Krzysztof Lis, Sina Honari, Pascal Fua, and Mathieu Salzmann. 2020. Detecting Road Obstacles by Erasing Them. ARXIV (2020)

    2020

  24. [33]

    Krzysztof Lis, Krishna Nakka, Pascal Fua, and Mathieu Salzmann. 2019. Detecting the unexpected via image resynthesis. In Proceedings of the IEEE International Conference on Computer Vision . 2152–2161

    2019

  25. [34]

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. 2024. Grounding DINO: Marry- ing DINO with grounded pre-training for open-set object detection. InProceedings of the European Conference on Computer Vis...

    2024

  26. [35]

    Yuyuan Liu, Choubo Ding, Yu Tian, Guansong Pang, Vasileios Belagiannis, Ian Reid, and Gustavo Carneiro. 2023. Residual pattern learning for pixel-wise out- of-distribution detection in semantic segmentation. In Proceedings of the IEEE International Conference on Computer Visio...

    2023

  27. [36]

    Ziyin Liu, Zhikang Wang, Paul Pu Liang, Russ R Salakhutdinov, Louis-Philippe Morency, and Masahito Ueda. 2019. Deep Gamblers: Learning to Abstain with Portfolio Theory. In Proceedings of the Advances in Neural Information Processing Systems, Vol. 32

    2019

  28. [37]

    Jonathan Long, Evan Shelhamer, and Trevor Darrell. 2015. Fully Convolutional Networks for Semantic Segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 3431–3440

    2015

  29. [38]

    Ilya Loshchilov and Frank Hutter. 2018. Decoupled weight decay regularization. In Proceedings of the International Conference on Learning Representations

    2018

  30. [39]

    Dimity Miller, Niko Sünderhauf, Michael Milford, and Feras Dayoub. 2021. Uncer- tainty for identifying open-set errors in visual object detection. RAL 7, 1 (2021), 215–222

    2021

  31. [40]

    Nazir Nayal, Misra Yavuz, Joao F Henriques, and Fatma Güney. 2023. RbA: Seg- menting unknown regions rejected by all. In Proceedings of the IEEE International Conference on Computer Vision . 711–722

    2023

  32. [41]

    Nguyen, Jason Yosinski, and Jeff Clune

    Anh M. Nguyen, Jason Yosinski, and Jeff Clune. 2015. Deep neural networks are easily fooled: High confidence predictions for unrecognizable images. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    2015

  33. [42]

    Jian-Jun Qiao, Zhi-Qi Cheng, Xiao Wu, Wei Li, and Ji Zhang. 2022. Real-time Semantic Segmentation with Parallel Multiple Views Feature Augmentation. In Proceedings of the ACM International Conference on Multimedia . 6300–6308

    2022

  34. [43]

    Jian-Jun Qiao, Xiao Wu, Jun-Yan He, Wei Li, and Qiang Peng. 2022. SWNet: A Deep Learning Based Approach for Splashed Water Detection on Road. IEEE Transactions on Intelligent Transportation Systems 23, 4 (2022), 3012–3025

    2022

  35. [44]

    Shyam Nandan Rai, Fabio Cermelli, Barbara Caputo, and Carlo Masone. 2024. Mask2anomaly: Mask transformer for universal open-set segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    2024

  36. [45]

    Yu Tian, Yuyuan Liu, Guansong Pang, Fengbei Liu, Yuanhong Chen, and Gustavo Carneiro. 2022. Pixel-Wise Energy-Biased Abstention Learning for Anomaly Segmentation on Complex Urban Driving Scenes. In Proceedings of the European Conference on Computer Vision

    2022

  37. [46]

    Tomas Vojir, Tomáš Šipka, Rahaf Aljundi, Nikolay Chumerin, Daniel Olmeda Reino, and Jiri Matas. 2021. Road Anomaly Detection by Partial Image Recon- struction with Segmentation Coupling. In Proceedings of the IEEE International Conference on Computer Vision

    2021

  38. [47]

    Wei Wang, Zhiqiang Chen, Xu Tao, Yi Cao, Liang Cheng, and Cheng Deng. 2022. Multimodal Anomaly Detection via Contrastive Fusion. IEEE Transactions on Neural Networks and Learning Systems 33, 12 (2022), 7597–7610

    2022

  39. [48]

    Yingda Xia, Yi Zhang, Fengze Liu, Wei Shen, and Alan Yuille. 2020. Synthesize Then Compare: Detecting Failures and Anomalies for Semantic Segmentation. In Proceedings of the European Conference on Computer Vision

    2020

  40. [49]

    Lewei Yao, Jianyuan Han, Yizeng Wen, Xiaodan Liang, Dan Xu, Wei Zhang, Zhen Li, Chunjing Xu, and Hang Xu. 2022. Detclip: Dictionary-enriched visual-concept paralleled pre-training for open-world detection. ARXIV (2022). ACM MM’25, October 27–31, 2025, Dublin, Ireland Trovato et al

    2022

  41. [50]

    Fisher Yu, Haofeng Chen, Xin Wang, Wenqi Xian, Yingying Chen, Fangchen Liu, Vashisht Madhavan, and Trevor Darrell. 2020. BDD100K: A diverse driving dataset for heterogeneous multitask learning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    2020

  42. [51]

    Alireza Zareian, Kevin Dela Rosa, Dengke Hu, and Shih-Fu Chang. 2021. Open- vocabulary object detection using captions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 14393–14402

    2021

  43. [52]

    Ji Zhang, Xiao Wu, Zhi-Qi Cheng, Qi He, and Wei Li. 2023. Improving anomaly segmentation with multi-granularity cross-domain alignment. In Proceedings of the ACM International Conference on Multimedia . 8515–8524

    2023

  44. [53]

    Wenjie Zhao, Jia Li, Xin Dong, Yu Xiang, and Yunhui Guo. 2024. Segment every out-of-distribution object. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 3910–3920. Appendix In this appendix, we provide additional experimental results, evalu- ...

    2024

  45. [2019]

    In Proceedings of the IEEE International Conference on Computer Vision

    Expectation-Maximization Attention Networks for Semantic Segmentation. In Proceedings of the IEEE International Conference on Computer Vision

  46. [2021]

    In Proceedings of the IEEE International Conference on Computer Vision

    Standardized Max Logits: A Simple Yet Effective Approach for Identifying Unexpected Road Obstacles in Urban-Scene Segmentation. In Proceedings of the IEEE International Conference on Computer Vision . 15425–15434

  47. [2022]

    In Proceedings of the ACM International Conference on Multimedia

    Gsrformer: Grounded Situation Recognition Transformer with Alternate Se- mantic Attention Refinement. In Proceedings of the ACM International Conference on Multimedia. 3272–3281

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.