Research map / Label-efficient and robust learning / OOD and anomaly detection
Detecting Semantic Anomalies
Faruk Ahmed, Aaron Courville
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 rotation prediction as an auxiliary task (ablation) | CIFAR-10 hold-out-class classifier with MSP · CIFAR-10 hold-out-class | average precision: 34.92 ± 0.41 -> 40.22 ± 0.16 | Improved |
| Add rotation prediction as an auxiliary task (ablation) | CIFAR-10 hold-out-class classifier with ODIN · CIFAR-10 hold-out-class | average precision: 36.63 ± 0.61 -> 41.56 ± 0.15 | Improved |
| Add rotation prediction as an auxiliary task (ablation) | STL-10 hold-out-class classifier with MSP · STL-10 hold-out-class | average precision: 21.07 ± 0.25 -> 24.41 ± 0.23 | Improved |
| Add rotation prediction as an auxiliary task (ablation) | STL-10 hold-out-class classifier with ODIN · STL-10 hold-out-class | average precision: 22.27 ± 0.29 -> 25.14 ± 0.45 | Improved |
| Add rotation prediction as an auxiliary task (ablation) | CIFAR-10 classifier · CIFAR-10 hold-out-class | accuracy: 96.03 ± 0.00 -> 96.83 ± 0.02 | Improved |
| Add rotation prediction as an auxiliary task (ablation) | STL-10 classifier · STL-10 hold-out-class | accuracy: 85.77 ± 0.13 -> 89.92 ± 0.08 | Improved |
| Add rotation prediction as an auxiliary task (ablation) | ImageNet Dog subset classifier with MSP · ImageNet Dog subset | average precision: 23.92 ± 0.49 -> 24.66 ± 0.58 | No clear effect |
| Add rotation prediction as an auxiliary task (ablation) | ImageNet Car subset classifier with MSP · ImageNet Car subset | average precision: 21.54 ± 0.62 -> 21.66 ± 0.19 | No clear effect |
| Add CPC as an auxiliary task (ablation) | ImageNet subset classifier with MSP · ImageNet subsets | average precision: Dog 20.84 ± 0.50 -> 21.43 ± 0.63; Car 19.86 ± 0.21 -> 22.21 ± 0.44; Snake 18.20 ± 0.76 -> 18.78 ± 0.40; Spider 22.03 ± 0 | Mixed |
| Add CPC as an auxiliary task (ablation) | ImageNet subset classifier with ODIN · ImageNet subsets | average precision: Dog 22.77 ± 0.74 -> 24.08 ± 0.63; Car 21.42 ± 0.48 -> 23.61 ± 0.57; Snake 18.67 ± 1.07 -> 20.39 ± 0.60; Spider 24.08 ± 0 | Mixed |
| Randomly mask a 16 × 16 region within the central 21 × 21 region (ablation) | CIFAR-10 classifier · CIFAR-10 hold-out-class | average precision with MSP: 34.92 ± 0.41 -> 34.41 ± 0.74 | Hurt |
| Replace classification-only training with random-center masking (ablation) | CIFAR-10 classifier · CIFAR-10 hold-out-class | accuracy: 96.03 ± 0.00 -> 96.27 ± 0.05 | Improved |
Closest papers by meaning
- Open-Vocabulary Video Anomaly Detection (CVPR 2024 · similarity 0.88)
- ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms (ICLR 2024 · similarity 0.88)
- Generalize or Detect? Towards Robust Semantic Segmentation Under Multiple Distribution Shifts (NeurIPS 2024 · similarity 0.87)
- COUNTS: Benchmarking Object Detectors and Multimodal Large Language Models under Distribution Shifts (CVPR 2025 · similarity 0.87)
- RUNA: Object-Level Out-of-Distribution Detection via Regional Uncertainty Alignment of Multimodal Representations (AAAI 2025 · similarity 0.87)
- Diffusion-based Semantic Outlier Generation via Nuisance Awareness for Out-of-Distribution Detection (AAAI 2025 · similarity 0.87)
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.
