Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Bayesian Embeddings for Few-Shot Open World Recognition
John Willes, James Harrison, Ali Harakeh, Chelsea Finn, Marco Pavone, Steven L. Waslander
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 adaptation loss (ablation) | SC-FLOWR · TieredImageNet small-context | H-measure: 11.94 -> 13.29 | Improved |
| add test-time fine-tuning on top of adaptation loss (ablation) | SC-FLOWR · TieredImageNet small-context | H-measure: 13.29 -> 15.04 | Improved |
| add test-time fine-tuning on top of adaptation loss (ablation) | SC-FLOWR · TieredImageNet small-context | accuracy: 48.16 -> 51.64 | Improved |
| add adaptation loss (ablation) | SC-FLOWR · TieredImageNet small-context | accuracy: 46.24 -> 48.16 | Improved |
| use CRP class prior instead of uniform (ablation) | SC-FLOWR · TieredImageNet small-context | H-measure: 13.54 -> 15.04 | Improved |
| use CRP class prior instead of uniform (ablation) | SC-FLOWR · TieredImageNet small-context | incremental accuracy: 41.92 -> 39.39 | Hurt |
| add trace regularization beta=0.1 in pre-training (ablation) | Sup-E pre-training · MiniImageNet 64-way val | accuracy: 42.26 -> 44.44 | Improved |
| trace regularization beta=0.01 instead of none (ablation) | Sup-E pre-training · MiniImageNet 64-way val | accuracy: 42.26 -> 42.34 | No clear effect |
| use embedding pre-training (Sup-E) instead of FC head (Sup-FC) | SC-FLOWR · TieredImageNet small-context | H-measure: 13.31 -> 15.04 | Improved |
| use embedding pre-training (Sup-E) instead of FC head (Sup-FC) | NCM · MiniImageNet small-context | accuracy: 37.88 -> 27.49 | Hurt |
| use H-measure instead of AUROC to compare PEELER and EVM | PEELER vs EVM · TieredImageNet small-context | metric discrimination: AUROC 53.24 vs 54.96; H-measure 1.00 vs 2.36 | Improved |
| SC-FLOWR vs best baseline | SC-FLOWR · TieredImageNet small-context | H-measure: 2.36 -> 15.04 | Improved |
Closest papers by meaning
- Toward Few-Shot Learning in the Open World: A Review and Beyond (TPAMI 2025 · similarity 0.90)
- Scaling Few-Shot Learning for the Open World (AAAI 2024 · similarity 0.90)
- Open-World Objectness Modeling Unifies Novel Object Detection (CVPR 2025 · similarity 0.86)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.86)
- Few-Shot Recognition via Stage-Wise Retrieval-Augmented Finetuning (CVPR 2025 · similarity 0.86)
- Detecting Open World Objects via Partial Attribute Assignment (CVPR 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.
