mlsanityData to 13 Oct 2026

BUSSARD: Normalizing Flows for Bijective Universal Scene-Specific Anomalous Relationship Detection

arXiv 2603.16645 ↗First posted 17 Mar 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

We propose Bijective Universal Scene-Specific Anomalous Relationship Detection (BUSSARD), a normalizing flow-based model for detecting anomalous relations in scene graphs, generated from images. Our work follows a multimodal approach, embedding object and relationship tokens from scene graphs with a language model to leverage semantic knowledge from the real world. A normalizing flow model is used to learn bijective transformations that map object-relation-object triplets from scene graphs to a simple base distribution (typically Gaussian), allowing anomaly detection through likelihood estimation. We evaluate our approach on the SARD dataset containing office and dining room scenes. Our method achieves around 10% better AUROC results compared to the current state-of-the-art model, while simultaneously being five times faster. Through ablation studies, we demonstrate superior robustness and universality, particularly regarding the use of synonyms, with our model maintaining stable performance while the baseline shows 17.5% deviation. This work demonstrates the strong potential of learning-based methods for relationship anomaly detection in scene graphs. Our code is available at https://github.com/mschween/BUSSARD .

Key facts. No citing papers in the map yet (data through 13 Oct 2026). Builds on 52 of its 76 references that are in the map. First posted on arXiv 17 Mar 2026.

Builds on

52 of 76 references.

References · 52

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,950
  2. 2020End-to-End Object Detection with Transformerscited by 7,396
  3. 2013Efficient Estimation of Word Representations in Vector Spacecited by 4,360
  4. 2016Visual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotationscited by 2,649
  5. 2015Variational Inference with Normalizing Flowscited by 2,163
  6. 2016Density estimation using Real NVPcited by 2,033
  7. 2018Glow: Generative Flow with Invertible 1x1 Convolutionscited by 1,542
  8. 2020MiniLM: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformerscited by 1,435
  9. 2016ConceptNet 5.5: An Open Multilingual Graph of General Knowledgecited by 1,065
  10. 2014NICE: Non-linear Independent Components Estimationcited by 1,054
  11. 2020MPNet: Masked and Permuted Pre-training for Language Understandingcited by 976
  12. 2021Towards Total Recall in Industrial Anomaly Detectioncited by 774
38 more
  1. 2017Masked Autoregressive Flow for Density Estimationcited by 680
  2. 2018Real-world Anomaly Detection in Surveillance Videoscited by 573
  3. 2022SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentationcited by 399
  4. 2023WinCLIP: Zero-/Few-Shot Anomaly Classification and Segmentationcited by 314
  5. 2023AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detectioncited by 271
  6. 2020Same Same But DifferNet: Semi-Supervised Defect Detection with Normalizing Flowscited by 212
  7. 2021The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localizationcited by 177
  8. 2023AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Modelscited by 171
  9. 2023Multimodal Industrial Anomaly Detection via Hybrid Fusioncited by 138
  10. 2022Asymmetric Student-Teacher Networks for Industrial Anomaly Detectioncited by 116
  11. 2022RelTR: Relation Transformer for Scene Graph Generationcited by 95
  12. 2024Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detectioncited by 93
  13. 2023BMAD: Benchmarks for Medical Anomaly Detectioncited by 82
  14. 2022Normalizing Flows for Human Pose Anomaly Detectioncited by 59
  15. 2021Anomaly Detection in Dynamic Graphs via Transformercited by 58
  16. 2023Panoptic Video Scene Graph Generationcited by 46
  17. 2024EGTR: Extracting Graph from Transformer for Scene Graph Generationcited by 45
  18. 2024From Pixels to Graphs: Open-Vocabulary Scene Graph Generation with Vision-Language Modelscited by 43
  19. 2025The MVTec AD 2 Dataset: Advanced Scenarios for Unsupervised Anomaly Detectioncited by 29
  20. 2025LogicAD: Explainable Anomaly Detection via VLM-based Text Feature Extractioncited by 24
  21. 2025Towards Training-free Anomaly Detection with Vision and Language Foundation Modelscited by 24
  22. 2021Unconditional Scene Graph Generationcited by 22
  23. 2024OED: Towards One-stage End-to-End Dynamic Scene Graph Generationcited by 22
  24. 2023Unbiased Scene Graph Generation via Two-stage Causal Modelingcited by 21
  25. 2024An Exploratory Study on Human-Centric Video Anomaly Detection through Variational Autoencoders and Trajectory Predictioncited by 19
  26. 2023Compositional Feature Augmentation for Unbiased Scene Graph Generationcited by 15
  27. 2021Detecting Anomalies in Semantic Segmentation with Prototypescited by 13
  28. 2024CSAD: Unsupervised Component Segmentation for Logical Anomaly Detectioncited by 11
  29. 2025Universal Scene Graph Generationcited by 11
  30. 2023Environment-Invariant Curriculum Relation Learning for Fine-Grained Scene Graph Generationcited by 9
  31. 2025LLM Meets Scene Graph: Can Large Language Models Understand and Generate Scene Graphs? A Benchmark and Empirical Studycited by 9
  32. 2024A SAM-guided Two-stream Lightweight Model for Anomaly Detectioncited by 7
  33. 2025ComplexVAD: Detecting Interaction Anomalies in Videocited by 7
  34. 2025Learning 4D Panoptic Scene Graph Generation from Rich 2D Visual Scenecited by 6
  35. 2025Synthetic Visual Genomecited by 6
  36. 2023Comparison of Affine and Rational Quadratic Spline Coupling and Autoregressive Flows through Robust Statistical Testscited by 5
  37. 2025Multi-Flow: Multi-View-Enriched Normalizing Flows for Industrial Anomaly Detectioncited by 5
  38. 2021Ano-Graph: Learning Normal Scene Contextual Graphs to Detect Video Anomaliescited by 4

50 most cited of 52 All 52 (JSON)

Built on by

Not cited yet.