Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Bayesian Triplet Loss: Uncertainty Quantification in Image Retrieval
Frederik Warburg, Martin Jørgensen, Javier Civera, Søren Hauberg
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 Gaussian Bayesian triplet loss instead of triplet regression | Triplet loss, ResNet50 · CUB200 | ECE@5: 0.331 -> 0.037 | Improved |
| use Bayesian triplet loss instead of plain triplet loss | ResNet50 · CUB200 | Recall@1: 0.648 -> 0.612 | Hurt |
| use expected distance instead of Euclidean distance for nearest neighbor search | Bayes Triplet, ResNet50 · CUB200 | Recall@1: 0.612 -> 0.632 | Improved |
| use expected distance for search with Densenet161 | Bayes Triplet, Densenet161 · CUB200 | Recall@1: 0.683 -> 0.617 | Hurt |
| use larger Densenet161 backbone instead of ResNet50 | Triplet · CUB200 | Recall@1: 0.648 -> 0.717 | Improved |
| add MC dropout (MC=50) to triplet loss | Triplet, Densenet161 · CUB200 | Recall@1: 0.717 -> 0.349 | Hurt |
| add MC dropout (MC=50) to triplet loss | Triplet, Densenet161 · MSLS | Recall@1: 0.386 -> 0.282 | Hurt |
| use von Mises-Fisher embeddings instead of triplet loss | ResNet50 · Car-196 (OOD) | Recall@1: 0.451 -> 0.471 | Improved |
| use Gaussian Bayesian triplet loss instead of triplet regression | Densenet161 · Car-196 (OOD) | ECE@5: 0.456 -> 0.134 | Improved |
| use Gaussian Bayesian triplet loss instead of triplet regression | Densenet161 · MSLS | ECE@5: 0.543 -> 0.264 | Improved |
| use Gaussian Bayesian triplet loss instead of triplet loss | ResNet50 · MSLS | Recall@1: 0.350 -> 0.354 | No clear effect |
Closest papers by meaning
- Risk Controlled Image Retrieval (AAAI 2025 · similarity 0.87)
- Benchmarking Uncertainty Disentanglement: Specialized Uncertainties for Specialized Tasks (NeurIPS 2024 · similarity 0.87)
- Composed Image Retrieval with Text Feedback via Multi-grained Uncertainty Regularization (ICLR 2024 · similarity 0.86)
- Learning with Noisy Triplet Correspondence for Composed Image Retrieval (CVPR 2025 · similarity 0.86)
- REPEAT: Improving Uncertainty Estimation in Representation Learning Explainability (AAAI 2025 · similarity 0.85)
- From Risk to Uncertainty: Generating Predictive Uncertainty Measures via Bayesian Estimation (ICLR 2025 · similarity 0.84)
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.
