Research map / Detection and segmentation / Open-set and few-shot detection
Semantic-Guided Multi-Attention Localization for Zero-Shot Learning
Yizhe Zhu, Jianwen Xie, Zhiqiang Tang, Xi Peng, Ahmed Elgammal
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 localized part patches to original image (ablation) | SGMA zero-shot model · ZSL PS average | mean class accuracy (%): 59.8 -> 65.2 | Improved |
| use parts only without original image (ablation) | SGMA zero-shot model · ZSL PS average | mean class accuracy (%): 59.8 -> 52.1 | Hurt |
| add random crops instead of localized parts (ablation) | SGMA zero-shot model · ZSL PS average | mean class accuracy (%): 59.8 -> 57.5 | Hurt |
| add class-center triplet loss and combine both inference branches (ablation) | SGMA zero-shot model · ZSL PS average | mean class accuracy (%): 59.8 -> 63.7 | Improved |
| class-center triplet inference instead of embedding softmax inference (ablation) | SGMA zero-shot model · ZSL PS average | mean class accuracy (%): 60.2 -> 62.6 | Improved |
| add class-center triplet loss, softmax-branch inference only (ablation) | SGMA zero-shot model · ZSL PS average | mean class accuracy (%): 59.8 -> 60.2 | Improved |
| add multi-attention loss (compactness and diversity) (ablation) | SGMA zero-shot model · CUB part detection | average precision (%): 47.6 -> 61.5 | Improved |
| share CNN weights between image and part patches (ablation) | SGMA zero-shot model · CUB PS | mean class accuracy (%): -3.6 | Hurt |
| share CNN weights between image and part patches (ablation) | SGMA zero-shot model · AWA PS | mean class accuracy (%): -2.7 | Hurt |
| increase number of attention parts beyond 2 (ablation) | SGMA zero-shot model · CUB, AWA, FLO | mean class accuracy (%) | No clear effect |
| SGMA instead of LDF in generalized ZSL | SGMA zero-shot model · CUB GZSL | harmonic mean H (%): 39.9 -> 48.5 | Improved |
| SGMA vs LDF on seen-class accuracy | SGMA zero-shot model · CUB GZSL | A_S (%): 81.6 -> 71.3 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- SPDA-CNN — SPDA-CNN: Unifying Semantic Part Detection and Abstraction for Fine-Grained Recognition (CVPR 2016) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Semantic-guided Reinforced Region Embedding for Generalized Zero-Shot Learning (AAAI 2021 · similarity 0.91)
- Predicting Visual Exemplars of Unseen Classes for Zero-Shot Learning (ICCV 2017 · similarity 0.90)
- Discriminative Region-Based Multi-Label Zero-Shot Learning (ICCV 2021 · similarity 0.89)
- TransZero: Attribute-Guided Transformer for Zero-Shot Learning (AAAI 2022 · similarity 0.89)
- ZeroMamba: Exploring Visual State Space Model for Zero-Shot Learning (AAAI 2025 · similarity 0.89)
- Epsilon: Exploring Comprehensive Visual-Semantic Projection for Multi-Label Zero-Shot Learning (AAAI 2025 · 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.
