Research map / Label-efficient and robust learning / Domain adaptation and generalization
UM-Adapt: Unsupervised Multi-Task Adaptation Using Adversarial Cross-Task Distillation
Jogendra Nath Kundu, Nishank Lakkakula, R. Venkatesh Babu
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 CCR instead of FCF content regularizer (ablation) | UM-Adapt-B · NYUD-v2 depth | rel error: 0.218 -> 0.192 | Mixed |
| use CCR instead of FCF content regularizer (ablation) | UM-Adapt-B · NYUD-v2 segmentation | mean IoU: 0.154 -> 0.163 | Improved |
| use CCR instead of FCF content regularizer (ablation) | UM-Adapt-B · NYUD-v2 surface normals | mean angle error: 24.6 -> 23.8 | Improved |
| add cross-task distillation from frozen task-transfer networks (ablation) | UM-Adapt-B(CCR) · NYUD-v2 segmentation | mean IoU: 0.163 -> 0.214 | Improved |
| add cross-task distillation from frozen task-transfer networks (ablation) | UM-Adapt-B(CCR) · NYUD-v2 depth | rel error: 0.192 -> 0.178 | Improved |
| use two task-transfer networks instead of one (ablation) | UM-Adapt-(noAdv.) · NYUD-v2 segmentation | mean IoU: 0.189 -> 0.214 | Improved |
| use two task-transfer networks instead of one (ablation) | UM-Adapt-(noAdv.) · NYUD-v2 depth | rel error: 0.181 -> 0.178 | Improved |
| add adversarial update of task-transfer networks (ablation) | UM-Adapt-(noAdv.) · NYUD-v2 segmentation | mean IoU: 0.214 -> 0.221 | Improved |
| add adversarial update of task-transfer networks (ablation) | UM-Adapt-(noAdv.) · Cityscapes segmentation | mean IoU: 0.408 -> 0.420 | Improved |
| add adversarial update of task-transfer networks (ablation) | UM-Adapt-(noAdv.) · KITTI depth | rel error: 0.28 -> 0.27 | Improved |
| add adversarial update of task-transfer networks (ablation) | UM-Adapt-(noAdv.) · training stability | stability | Hurt |
| add multi-task adaptation (UM-Adapt-(Adv.)) to source-only baseline | Multi-task baseline · NYUD-v2 segmentation | mean IoU: 0.022 -> 0.221 | Improved |
Closest papers by meaning
- Transferable Representation Learning with Deep Adaptation Networks (TPAMI 2018 · similarity 0.88)
- Learning Across Tasks and Domains (ICCV 2019 · similarity 0.88)
- Unsupervised Domain Adaptation with Residual Transfer Networks (NeurIPS 2016 · similarity 0.87)
- Unsupervised Visual Domain Adaptation: A Deep Max-Margin Gaussian Process Approach (CVPR 2019 · similarity 0.87)
- Co-regularized Alignment for Unsupervised Domain Adaptation (NeurIPS 2018 · similarity 0.87)
- A General Representation-Based Approach to Multi-Source Domain Adaptation (ICML 2025 · similarity 0.86)
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.
