Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Semi-transductive Learning for Generalized Zero-Shot Sketch-Based Image Retrieval
Ce Ge, Jingyu Wang, Qi Qi, Haifeng Sun, Tong Xu, Jianxin Liao
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 |
|---|---|---|---|
| Use semi-transductive learning with unlabeled unseen-class images | STL · Sketchy | mAP (%): 44.5 -> 53.0 | Improved |
| Use semi-transductive learning with unlabeled unseen-class images | STL · TU-Berlin | mAP (%): 31.2 -> 40.2 | Improved |
| Add discriminative semantic alignment to ranking learning (ablation) | STL · Sketchy | mAP (%): 16.3 -> 27.4 | Improved |
| Add discriminative semantic alignment to ranking learning (ablation) | STL · TU-Berlin | mAP (%): 9.6 -> 13.8 | Improved |
| Add distribution calibration using unlabeled unseen-class images and semi-transductive ranking (ablation) | STL with semantic alignment · Sketchy | mAP (%): 27.4 -> 53.0 | Improved |
| Add distribution calibration using unlabeled unseen-class images and semi-transductive ranking (ablation) | STL with semantic alignment · TU-Berlin | mAP (%): 13.8 -> 40.2 | Improved |
| Use a semi-heterogeneous encoder instead of a heterogeneous encoder (ablation) | STL ranking model · Sketchy | mAP (%): 8.7 -> 16.3 | Improved |
| Use a semi-heterogeneous encoder instead of a heterogeneous encoder (ablation) | STL ranking model · TU-Berlin | mAP (%): 5.3 -> 9.6 | Improved |
| Use a semi-heterogeneous encoder instead of a Siamese encoder (ablation) | STL ranking model · Sketchy | mAP (%): 11.2 -> 16.3 | Improved |
| Use a semi-heterogeneous encoder instead of a Siamese encoder (ablation) | STL ranking model · TU-Berlin | mAP (%): 6.7 -> 9.6 | Improved |
| Compress the embedding to 64D rather than use the 1024D STL embedding (ablation) | STL · Sketchy | mAP (%): 53.0 -> 44.6 | Hurt |
| Compress the embedding to 64D rather than use the 1024D STL embedding (ablation) | STL · TU-Berlin | mAP (%): 40.2 -> 36.3 | Hurt |
Closest papers by meaning
- Transferable Coupled Network for Zero-Shot Sketch-Based Image Retrieval (TPAMI 2021 · similarity 0.93)
- A Zero-Shot Framework for Sketch based Image Retrieval (ECCV 2018 · similarity 0.93)
- Zero-Shot Sketch-Based Image Retrieval via Graph Convolution Network (AAAI 2020 · similarity 0.92)
- Doodle to Search: Practical Zero-Shot Sketch-Based Image Retrieval (CVPR 2019 · similarity 0.91)
- Semantic-Aware Knowledge Preservation for Zero-Shot Sketch-Based Image Retrieval (ICCV 2019 · similarity 0.91)
- Asymmetric Mutual Alignment for Unsupervised Zero-Shot Sketch-Based Image Retrieval (AAAI 2024 · similarity 0.89)
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.
