Research map / Detection and segmentation / Open-set and few-shot detection
Synthesize then Compare: Detecting Failures and Anomalies for Semantic Segmentation
Yingda Xia, Yi Zhang, Fengze Liu, Wei Shen, Alan L. Yuille
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 synthesized image from cGAN to the comparison network (ablation) | SynthCP failure detector (FCN-8) · Cityscapes pixel-level | AUPR-Error: 52.17 -> 55.53 | Improved |
| add synthesized image from cGAN to the comparison network (ablation) | SynthCP failure detector (Deeplab-v2) · Cityscapes pixel-level | AUPR-Error: 48.76 -> 49.99 | Improved |
| add synthesized image from cGAN (ablation) | SynthCP failure detector (FCN-8) · Cityscapes image-level | MAE: 13.25 -> 11.58 | Improved |
| add synthesized image from cGAN (ablation) | SynthCP failure detector (FCN-8) · Cityscapes image-level | Pearson correlation: 58.34 -> 64.63 | Improved |
| train image-level and pixel-level heads jointly instead of separately (ablation) | SynthCP · Cityscapes pixel-level, FCN-8 | AUPR-Error: 54.14 -> 55.53 | Improved |
| train heads jointly instead of separately (ablation) | SynthCP · Cityscapes image-level, FCN-8 | MAE: 11.58 -> 12.69 | Hurt |
| ensemble SynthCP with VAE alarm | SynthCP · MSD pancreatic tumor | Pearson correlation: 61.11 -> 67.97 | Improved |
| ensemble SynthCP with VAE alarm | SynthCP · MSD pancreatic tumor | MAE: 18.13 -> 15.19 | Improved |
| apply MSP post-processing with threshold t=0.999 (ablation) | SynthCP · StreetHazards | AUPR: 8.1 -> 9.3 | Improved |
| apply MSP post-processing with threshold t=0.999 (ablation) | SynthCP · StreetHazards | FPR95: 46.0 -> 28.4 | Improved |
| apply MSP post-processing with threshold t=0.999 (ablation) | SynthCP · StreetHazards | AUROC: 81.9 -> 88.5 | Improved |
| lower post-processing threshold from 0.999 to 0.8 (ablation) | SynthCP · StreetHazards | AUPR: 9.3 -> 7.4 | Hurt |
Closest papers by meaning
- Pixel-Wise Anomaly Detection in Complex Driving Scenes (CVPR 2021 · similarity 0.88)
- Few Shot Part Segmentation Reveals Compositional Logic for Industrial Anomaly Detection (AAAI 2024 · similarity 0.85)
- A Dataset for Semantic Segmentation in the Presence of Unknowns (CVPR 2025 · similarity 0.85)
- MetaUAS: Universal Anomaly Segmentation with One-Prompt Meta-Learning (NeurIPS 2024 · similarity 0.85)
- FAST: Foreground‑aware Diffusion with Accelerated Sampling Trajectory for Segmentation‑oriented Anomaly Synthesis (NeurIPS 2025 · similarity 0.84)
- SemFlow: Binding Semantic Segmentation and Image Synthesis via Rectified Flow (NeurIPS 2024 · similarity 0.84)
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.
