Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
On Zero-Shot Recognition of Generic Objects
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use a trivial nearest-training-class mapping instead of distributed semantic representations | Linear ZSL on the standard ImageNet benchmark · ImageNet 1-hop | top-1 accuracy: 26.6 -> 20.27 | Hurt |
| Use a trivial nearest-training-class mapping instead of distributed semantic representations | Linear ZSL on the standard ImageNet benchmark · ImageNet 2-hops | top-1 accuracy: 6.3 -> 3.59 | Hurt |
| Use a trivial nearest-training-class mapping instead of distributed semantic representations | Linear ZSL on the standard ImageNet benchmark · ImageNet all | top-1 accuracy: 3.0 -> 1.53 | Hurt |
| Evaluate on a benchmark with maximal structural ratio and curated semantic and visual samples | Existing ZSL models · ImageNet 1-hop | top-1 error rate: 86% to 61% | Helped |
| Evaluate on a benchmark with maximal structural ratio and curated semantic and visual samples | GCN · ImageNet 1-hop (generalized ZSL) | top-1 error rate: 90% to 47% | Helped |
| Use test classes with more frequent label words | Linear ZSL baseline · 100-class ImageNet test splits | top-1 accuracy: r=0.89 | Helped |
| Use test classes with more frequent label words | GCN · 100-class ImageNet test splits | top-1 accuracy: r=0.74 | Helped |
| Select primary rather than secondary meanings for polysemous labels | ZSL models · 100-class ImageNet test splits | top-1 accuracy | Helped |
| Use test classes with larger sample populations | Linear and GCN ZSL models · ImageNet test splits | top-1 accuracy | Helped |
| Increase test-split structural ratio | Trivial similarity-based solution · 100-class ImageNet test splits | top-1 accuracy | Hurt |
| Increase test-split structural ratio | GCN and Linear ZSL models · 100-class ImageNet test splits | top-1 accuracy | Mixed |
| Replace the standard benchmark with the proposed curated benchmark | GCN-based models relative to linear baselines · Proposed 500-class benchmark | top-1 accuracy | Hurt |
Closest papers by meaning
- Context-Aware Zero-Shot Learning for Object Recognition (ICML 2019 · similarity 0.91)
- Zero-Shot Learning - the Good, the Bad and the Ugly (CVPR 2017 · similarity 0.90)
- TPR: Topology-Preserving Reservoirs for Generalized Zero-Shot Learning (NeurIPS 2024 · similarity 0.89)
- Data Distribution Distilled Generative Model for Generalized Zero-Shot Recognition (AAAI 2024 · similarity 0.88)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.88)
- S3A: Towards Realistic Zero-Shot Classification via Self Structural Semantic Alignment (AAAI 2024 · similarity 0.88)
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.