Research map / Architectures and efficiency / Attention and transformers
DeMT: Deformable Mixer Transformer for Multi-Task Learning of Dense Prediction
Yangyang Xu, Yibo Yang, Lefei Zhang
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 DeMT instead of MQTransformer | Swin-T multi-task model · NYUD-v2 | SemSeg IoU: 43.61 -> 46.36 | Improved |
| Use DeMT instead of MQTransformer | Swin-T multi-task model · PASCAL-Context | SemSeg IoU: 68.24 -> 69.71 | Improved |
| Add the deformable mixer (ablation) | DeMT baseline · NYUD-v2 | SemSeg IoU: 38.78 -> 42.40 | Improved |
| Add the task interaction block (ablation) | DeMT with deformable mixer · NYUD-v2 | SemSeg IoU: 42.40 -> 44.44 | Improved |
| Add the task query block (ablation) | DeMT with deformable mixer and task interaction · NYUD-v2 | SemSeg IoU: 44.44 -> 46.36 | Improved |
| Increase deformable mixer depth from 1 to 2 (ablation) | DeMT · NYUD-v2 | SemSeg IoU: 46.36 -> 46.90 | Improved |
| Increase deformable mixer depth from 2 to 4 (ablation) | DeMT · NYUD-v2 | SemSeg IoU: 46.90 -> 47.71 | Improved |
| Increase deformable mixer depth from 4 to 8 (ablation) | DeMT · NYUD-v2 | SemSeg IoU: 47.71 -> 47.16 | Hurt |
| Add 1/8-scale features to the 1/4-scale input (ablation) | DeMT · NYUD-v2 | SemSeg IoU: 7.51 -> 12.85 | Improved |
| Add 1/16-scale features (ablation) | DeMT using 1/4- and 1/8-scale features · NYUD-v2 | SemSeg IoU: 12.85 -> 40.32 | Improved |
| Add 1/32-scale features (ablation) | DeMT using three feature scales · NYUD-v2 | SemSeg IoU: 40.32 -> 46.36 | Improved |
Closest papers by meaning
- Multi-Task Learning for Dense Prediction Tasks: A Survey (TPAMI 2021 · similarity 0.88)
- Multi-Task Dense Prediction via Mixture of Low-Rank Experts (CVPR 2024 · similarity 0.85)
- Efficient Multitask Dense Predictor via Binarization (CVPR 2024 · similarity 0.85)
- TADFormer: Task-Adaptive Dynamic TransFormer for Efficient Multi-Task Learning (CVPR 2025 · similarity 0.84)
- Transforming Vision Transformer: Towards Efficient Multi-Task Asynchronous Learner (NeurIPS 2024 · similarity 0.83)
- ViT-CoMer: Vision Transformer with Convolutional Multi-scale Feature Interaction for Dense Predictions (CVPR 2024 · similarity 0.83)
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.
