Research map / Label-efficient and robust learning / Recognition and long-tail
Balanced MSE for Imbalanced Visual Regression
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace ordinary MSE with Balanced MSE | Visual regression · IMDB-WIKI-DIR | bMAE (All): 13.09 -> 12.66 | Helped |
| Replace ordinary MSE with Balanced MSE | Visual regression · IMDB-WIKI-DIR | MAE (All): 7.79 -> 8.12 | Hurt |
| Apply Balanced MSE to depth regression | NYUD2-DIR depth estimator · NYUD2-DIR | RMSE (All): 1.477 -> 1.251 | Helped |
| Apply Balanced MSE to depth regression | NYUD2-DIR depth estimator · NYUD2-DIR | δ1 (All): 0.677 -> 0.702 | Helped |
| Apply Balanced MSE to mesh regression | SPIN regressor · IHMR | bMPVPE (All): 116.1 -> 112.7 | Helped |
| Apply Balanced MSE to mesh regression | SPIN regressor · IHMR | bMPVPE (Tail 5%): 126.4 -> 128.1 | Hurt |
| Use Balanced MSE rather than vanilla least squares | Synthetic linear regressor · Synthetic regression, high-skew Normal | MSE: 5.521 -> 0.031 | Helped |
| Use Balanced MSE rather than inverse reweighting | Synthetic linear regressor · Synthetic regression, high-skew Exponential | MSE: 4.676 -> 0.001 | Helped |
| Use batch-based Monte Carlo instead of analytical GMM integration | Balanced MSE · Synthetic regression, high-skew MVN | MSE: 0.122 -> 0.126 | No clear effect |
| Fix noise scale at 6 rather than jointly learn it (ablation) | Balanced MSE (GAI) · IMDB-WIKI-DIR | bMAE (All): 12.66 -> 12.85 | Hurt |
| Fix noise scale at 7 rather than jointly learn it (ablation) | Balanced MSE (GAI) · IMDB-WIKI-DIR | bMAE (All): 12.67 -> 12.66 | No clear effect |
| Fix noise scale at 8 rather than jointly learn it (ablation) | Balanced MSE (GAI) · IMDB-WIKI-DIR | bMAE (All): 12.68 -> 12.66 | No clear effect |
Closest papers by meaning
- Deep Imbalanced Regression via Hierarchical Classification Adjustment (CVPR 2024 · similarity 0.88)
- Robust Visual Recognition with Class-Imbalanced Open-World Noisy Data (AAAI 2024 · similarity 0.86)
- Rethinking Imbalance in Image Super-Resolution for Efficient Inference (NeurIPS 2024 · similarity 0.85)
- Improve Representation for Imbalanced Regression through Geometric Constraints (CVPR 2025 · similarity 0.85)
- ConR: Contrastive Regularizer for Deep Imbalanced Regression (ICLR 2024 · similarity 0.85)
- Dist Loss: Enhancing Regression in Few-Shot Region through Distribution Distance Constraint (ICLR 2025 · similarity 0.85)
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.