Research map / Detection and segmentation / 3D detection and driving
Taskology: Utilizing Task Relations at Scale
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 |
|---|---|---|---|
| jointly train depth, motion, and segmentation with consistency constraints | Taskology · Cityscapes | Abs. Rel: 0.165 -> 0.125 | Helped |
| jointly train depth, motion, and segmentation with consistency constraints | Taskology · Cityscapes | mIoU: 0.455 -> 0.478 | Helped |
| add segmentation predictions as supervision for depth and motion (ablation) | Taskology depth & motion · Cityscapes | Abs. Rel: 0.165 -> 0.129 | Helped |
| add depth and motion predictions as consistency supervision for segmentation (ablation) | Taskology segmentation · Cityscapes | mIoU: 0.455 -> 0.471 | Helped |
| add motion-based consistency loss | PointPillar detector · Waymo Open Dataset, 5% labels | 3D mAP: 17.6 -> 23.5 | Helped |
| add motion-based consistency loss | PointPillar detector · Waymo Open Dataset, 5% labels | BEV mAP: 44.3 -> 51.1 | Helped |
| add motion-based consistency loss | PointPillar detector · Waymo Open Dataset, 20% labels | 3D mAPH: 16.4 -> 19.1 | Helped |
| add motion-based consistency loss | PointPillar detector · Waymo Open Dataset, 100% labels | BEV mAP: 75.0 -> 75.0 | No clear effect |
| add depth-normal consistency loss on unlabeled real ScanNet data | Depth and surface-normal models · SceneNet to ScanNet | Normal accuracy <11.25°: 9.2 -> 13.6 | Helped |
| add depth-normal consistency loss on unlabeled real ScanNet data | Depth and surface-normal models · SceneNet to ScanNet | Abs. Rel: 28.2 -> 24.9 | Helped |
| allow peer predictions to be up to 20 minutes stale (ablation) | Distributed depth and egomotion training · KITTI | Abs. Rel: about 0.143 | No clear effect |
Closest papers by meaning
- Building a Strong Pre-Training Baseline for Universal 3D Large-Scale Perception (CVPR 2024 · similarity 0.86)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 2024 · similarity 0.86)
- Going Beyond Multi-Task Dense Prediction with Synergy Embedding Models (CVPR 2024 · similarity 0.86)
- Towards Principled Task Grouping for Multi-Task Learning (NeurIPS 2026 · similarity 0.85)
- VMT-Adapter: Parameter-Efficient Transfer Learning for Multi-Task Dense Scene Understanding (AAAI 2024 · similarity 0.85)
- Multi-task Learning with 3D-Aware Regularization (ICLR 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.