Research map / Label-efficient and robust learning / Self-supervised and contrastive
Unsupervised Deep Metric Learning with Transformed Attention Consistency and Contrastive Clustering Loss
Yang Li, Shichao Kan, Zhihai He
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 contrastive clustering loss (ablation) | TAC-CCL baseline · CUB-200-2011 | Recall@1 (%): 53.9 -> 55.7 | Improved |
| add contrastive clustering loss (ablation) | TAC-CCL baseline · Cars-196 | Recall@1 (%): 43.0 -> 44.7 | Improved |
| add contrastive clustering loss (ablation) | TAC-CCL baseline · Stanford Online Products | Recall@1 (%): 62.5 -> 63.0 | Improved |
| add transformed attention consistency on top of CCL (ablation) | TAC-CCL · CUB-200-2011 | Recall@1 (%): 55.7 -> 57.5 | Improved |
| add transformed attention consistency on top of CCL (ablation) | TAC-CCL · Cars-196 | Recall@1 (%): 44.7 -> 46.1 | Improved |
| add transformed attention consistency on top of CCL (ablation) | TAC-CCL · Stanford Online Products | Recall@1 (%): 63.0 -> 63.9 | Improved |
| add contrastive clustering loss (ablation) | TAC-CCL baseline · Cars-196 | Recall@8 (%): 76.0 -> 75.7 | Hurt |
| add contrastive clustering loss (ablation) | TAC-CCL baseline · Stanford Online Products | Recall@100 (%): 87.2 -> 87.2 | No clear effect |
| increase number of clusters K beyond 100 (50, 100, 200, 500, 1000) (ablation) | TAC-CCL · CUB-200-2011 | Recall@1 (%) | Hurt |
| increase embedding size from 64 up to 1024 (ablation) | TAC-CCL · CUB-200-2011 | Recall@1 (%) | Improved |
| use TAC-CCL over the best prior unsupervised method (Instance) | TAC-CCL · CUB-200-2011 | Recall@1 (%): 46.2 -> 57.5 | Improved |
| use TAC-CCL over the best prior unsupervised method (Instance) | TAC-CCL · Cars-196 | Recall@1 (%): 41.3 -> 46.1 | Improved |
Closest papers by meaning
- Compressed Self-Attention for Deep Metric Learning (AAAI 2020 · similarity 0.89)
- Cross-Image-Attention for Conditional Embeddings in Deep Metric Learning (CVPR 2023 · similarity 0.89)
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric (CVPR 2018 · similarity 0.88)
- Beyond Supervised vs. Unsupervised: Representative Benchmarking and Analysis of Image Representation Learning (CVPR 2022 · similarity 0.88)
- MIC: Mining Interclass Characteristics for Improved Metric Learning (ICCV 2019 · similarity 0.87)
- Sharing Matters for Generalization in Deep Metric Learning (TPAMI 2020 · similarity 0.87)
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.
