Research map / Detection and segmentation / 3D detection and driving
Visual Exemplar Driven Task-Prompting for Unified Perception in Autonomous Driving
Read the paperCheck 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 |
|---|---|---|---|
| add visual-exemplar-driven task-specific prompts (ablation) | VE-Prompt with shared transformer encoder · BDD100K Disjoint-balance | mAP: 32.2 -> 33.9 | Helped |
| add visual-exemplar-driven task-specific prompts (ablation) | VE-Prompt with shared transformer encoder · BDD100K Disjoint-balance | mIoU (SS): 60.5 -> 61.2 | Helped |
| add visual-exemplar-driven task-specific prompts (ablation) | VE-Prompt with shared transformer encoder · BDD100K Disjoint-balance | mIoU (DA): 86.5 -> 87.4 | Helped |
| add a shared transformer encoder (ablation) | DINO-based multi-task model · BDD100K Disjoint-balance | mIoU (SS): 58.1 -> 60.5 | Helped |
| use pre-head prompting instead of post-head prompting (ablation) | VE-Prompt · BDD100K Disjoint-balance | mIoU (SS): 58.9 -> 61.2 | Helped |
| use pre-head prompting instead of post-head prompting (ablation) | VE-Prompt · BDD100K Disjoint-balance | mIoU (DA): 86.4 -> 87.4 | Helped |
| initialize prompts with CLIP instead of random initialization (ablation) | VE-Prompt · BDD100K Disjoint-balance | mAP: 33.5 -> 33.9 | Helped |
| make task-specific prompts trainable rather than fixed (ablation) | VE-Prompt · BDD100K Disjoint-balance | mAP: 33.3 -> 33.9 | Helped |
| increase the number of visual exemplars to 5 (ablation) | VE-Prompt · BDD100K Disjoint-balance | mAP | Helped |
| use pseudo labeling instead of zeroing missing-label losses | multi-task baseline · BDD100K Full | mIoU (SS): 58.6 -> 60.9 | Helped |
| use pseudo labeling instead of zeroing missing-label losses | multi-task baseline · BDD100K Full | mIoU (DA): 89.3 -> 89.3 | No clear effect |
| use Round-robin task scheduling instead of zeroing loss | multi-task baseline · BDD100K Disjoint-normal | mIoU (SS): 55.7 -> 61.0 | Helped |
Closest papers by meaning
- M3Net: Multimodal Multi-task Learning for 3D Detection, Segmentation, and Occupancy Prediction in Autonomous Driving (AAAI 2025 · similarity 0.90)
- MMTL-UniAD: A Unified Framework for Multimodal and Multi-Task Learning in Assistive Driving Perception (CVPR 2025 · similarity 0.87)
- Language Prompt for Autonomous Driving (AAAI 2025 · similarity 0.87)
- Prune2Drive: A Plug-and-Play Framework for Accelerating Vision-Language Models in Autonomous Driving (CVPR 2026 · similarity 0.87)
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 2026 · similarity 0.86)
- Knowledge Distillation from Single-Task Teachers to Multi-Task Student for End-to-End Autonomous Driving (AAAI 2024 · 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.