Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Semantically Tied Paired Cycle Consistency for Any-Shot Sketch-Based Image Retrieval
Anjan Dutta, Zeynep Akata
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 cycle consistency loss to adversarial-only training (ablation) | SEM-PCYC · Sketchy (Extended) zero-shot | mAP@all: 0.128 -> 0.147 | Improved |
| add classification loss to adversarial-only training (ablation) | SEM-PCYC · Sketchy (Extended) zero-shot | mAP@all: 0.128 -> 0.140 | Improved |
| add cycle consistency and classification losses together (adversarial on sketch and image) (ablation) | SEM-PCYC · Sketchy (Extended) zero-shot | mAP@all: 0.128 -> 0.213 | Improved |
| add adversarial loss for the semantic domain (ablation) | SEM-PCYC · Sketchy (Extended) zero-shot | mAP@all: 0.213 -> 0.349 | Improved |
| add the l2,1-regularized auto-encoder regularizer for side information (ablation) | SEM-PCYC · Sketchy (Extended) zero-shot | mAP@all: 0.323 -> 0.349 | Improved |
| add auto-encoder side-information selection (compress to 64d) (ablation) | SEM-PCYC · Sketchy (Extended) zero-shot | mAP@all: 0.382 -> 0.349 | Hurt |
| add auto-encoder side-information selection (compress to 64d) (ablation) | SEM-PCYC · QuickDraw (Extended) zero-shot | mAP@all: 0.185 -> 0.177 | Hurt |
| swap VGG-16 features for SE-ResNet-50 (FastText, 64d) (ablation) | SEM-PCYC · Sketchy (Extended) zero-shot | mAP@all: 0.344 -> 0.368 | Improved |
| swap VGG-16 features for SE-ResNet-50 (FastText, 64d) (ablation) | SEM-PCYC · TU-Berlin (Extended) zero-shot | mAP@all: 0.262 -> 0.349 | Improved |
| combine Word2Vec text embedding with Jiang-Conrath hierarchical embedding (ablation) | SEM-PCYC · Sketchy (Extended) zero-shot, 64d | mAP@all: 0.330 -> 0.349 | Improved |
| remove 5% to 30% of side-information dimensions (ablation) | SEM-PCYC · zero-shot SBIR | mAP@all: 5% to 30% | Improved |
| remove more than 20% to 40% of side information (ablation) | SEM-PCYC · zero-shot SBIR | mAP@all: beyond 20% to 40% | Hurt |
Closest papers by meaning
- Semantically Tied Paired Cycle Consistency for Zero-Shot Sketch-Based Image Retrieval (CVPR 2019 · similarity 0.97)
- 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)
- Transferable Coupled Network for Zero-Shot Sketch-Based Image Retrieval (TPAMI 2021 · similarity 0.91)
- Asymmetric Mutual Alignment for Unsupervised Zero-Shot Sketch-Based Image Retrieval (AAAI 2024 · similarity 0.89)
- Sketch Down the FLOPs: Towards Efficient Networks for Human Sketch (CVPR 2025 · 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.
