Research map / Detection and segmentation / Open-set and few-shot detection
BUSSARD: Normalizing Flows for Bijective Universal Scene-Specific Anomalous Relationship Detection
Melissa Schween, Mathis Kruse, Bodo Rosenhahn
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add autoencoder before the flow (ablation) | BUSSARD · SARD Dining Room | AUROC: 85.44 -> 97.85 | Improved |
| add autoencoder before the flow (ablation) | BUSSARD · SARD Office | AUROC: 85.74 -> 95.29 | Improved |
| add relation embedding (not objects only) (ablation) | BUSSARD · SARD Dining Room | AUROC: 95.20 -> 97.85 | Improved |
| add relation embedding (not objects only) (ablation) | BUSSARD · SARD Office | AUROC: 90.89 -> 95.29 | Improved |
| concatenate triplet vectors instead of summing (ablation) | BUSSARD · SARD Dining Room | AUROC: 91.22 -> 97.85 | Improved |
| concatenate triplet vectors instead of elementwise multiplication (ablation) | BUSSARD · SARD Dining Room | AUROC: 96.48 -> 97.85 | Improved |
| concatenate triplet vectors instead of summing (ablation) | BUSSARD · SARD Office | AUROC: 94.18 -> 95.29 | Improved |
| latent dimension 512 rather than 900 (input dim) (ablation) | BUSSARD · SARD Dining Room | AUROC: 85.44 -> 97.85 | Improved |
| replace EGTR with RelTR scene graph generator (ablation) | BUSSARD · SARD Dining Room | AUROC: 97.85 -> 95.57 | Hurt |
| replace EGTR with RelTR scene graph generator (ablation) | BUSSARD · SARD Office | AUROC: 95.29 -> 96.09 | No clear effect |
| replace GloVe with Qwen3-Embedding-8B (ablation) | BUSSARD · SARD Dining Room | AUROC: 97.85 -> 97.99 | No clear effect |
| add Gaussian noise sigma=0.10 to embeddings (ablation) | BUSSARD · SARD Office | AUROC: 95.29 -> 92.91 | Hurt |
Closest papers by meaning
- Scene-Graph ViT: End-to-End Open-Vocabulary Visual Relationship Detection (ECCV 2024 · similarity 0.87)
- Graphical Contrastive Losses for Scene Graph Parsing (CVPR 2019 · similarity 0.86)
- Tensorize, Factorize and Regularize: Robust Visual Relationship Learning (CVPR 2018 · similarity 0.86)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.86)
- UniVAD: A Training-free Unified Model for Few-shot Visual Anomaly Detection (CVPR 2025 · similarity 0.85)
- Open-Set Graph Anomaly Detection via Normal Structure Regularisation (ICLR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
