Research map / Label-efficient and robust learning / Semi-supervised learning
Learning Multiple Dense Prediction Tasks From Partially Annotated Data
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 |
|---|---|---|---|
| Use joint pairwise task-space consistency with conditional mappings and image-feature regularization instead of vanilla partial-label training | Multi-task dense prediction · Cityscapes one-label | Seg. (IoU): 69.50 -> 74.90 | Helped |
| Use joint pairwise task-space consistency with conditional mappings and image-feature regularization instead of vanilla partial-label training | Multi-task dense prediction · NYU-v2 random-label | Depth (aErr): 0.6624 -> 0.5787 | Helped |
| Use joint pairwise task-space consistency with conditional mappings and image-feature regularization instead of vanilla partial-label training | Multi-task dense prediction · PASCAL-Context one-label | Seg. (IoU): 48.0 -> 49.5 | Helped |
| Add task-pair conditioning to the mapping (ablation) | Ours · NYU-v2 random-label | Seg. (IoU): 34.13 -> 34.26 | Helped |
| Add task-pair conditioning to the mapping (ablation) | Ours · NYU-v2 one-label | Seg. (IoU): 29.19 -> 30.36 | Helped |
| Add image-feature regularization to the mapping (ablation) | Ours · NYU-v2 random-label | Seg. (IoU): 33.87 -> 34.26 | Helped |
| Add image-feature regularization to the mapping (ablation) | Ours · NYU-v2 one-label | Seg. (IoU): 28.36 -> 30.36 | Helped |
| Add semi-supervised perturbation consistency on top of cross-task consistency | Ours · NYU-v2 | performance | No clear effect |
| Use Direct-Map instead of the proposed pairwise latent mapping | Multi-task dense prediction · NYU-v2 one-label | Seg. (IoU): 30.36 -> 19.98 | Hurt |
| Use Discriminator-Loss instead of the proposed method | Multi-task dense prediction · Cityscapes one-label | Seg. (IoU): 69.50 -> 68.94 | Hurt |
| Add cross-task consistency to the semi-supervised setting | Multi-task model · PASCAL-Context random-label | Seg. (IoU): 59.0 -> 59.0 | No clear effect |
| Add Uncertainty loss weighting to Ours | Ours · NYU-v2 fully supervised | Norm. (mErr): 28.58 -> 26.78 | Helped |
Closest papers by meaning
- Joint-Task Regularization for Partially Labeled Multi-Task Learning (CVPR 2024 · similarity 0.88)
- Multi-Task Dense Prediction via Mixture of Low-Rank Experts (CVPR 2024 · similarity 0.87)
- DiffusionMTL: Learning Multi-Task Denoising Diffusion Model from Partially Annotated Data (CVPR 2024 · similarity 0.87)
- Distribution Matching for Multi-Task Learning of Classification Tasks: A Large-Scale Study on Faces & Beyond (AAAI 2024 · similarity 0.86)
- Efficient Multitask Dense Predictor via Binarization (CVPR 2024 · similarity 0.85)
- Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · 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.