Research map / Detection and segmentation / 3D detection and driving
ROAD: The Road Event Awareness Dataset for Autonomous Driving
Gurkirt Singh, Stephen Akrigg, Manuele Di Maio, Valentina Fontana, Reza Javanmard Alitappeh, Salman Khan, Suman Saha, Kossar Jeddisaravi, Farzad Yousefi, Jacob Culley, Tom Nicholson, Jordan Omokeowa and 4 more
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 |
|---|---|---|---|
| replace 2D ResNet50 backbone with I3D (ablation) | 3D-RetinaNet · ROAD test, frame-level events | frame-mAP (%): 12.8 -> 16.5 | Improved |
| replace I3D backbone with Slowfast (ablation) | 3D-RetinaNet · ROAD test, frame-level events | frame-mAP (%): 16.5 -> 23.7 | Improved |
| replace I3D backbone with Slowfast (ablation) | 3D-RetinaNet · ROAD test, frame-level actions | frame-mAP (%): 25.4 -> 31.1 | Improved |
| replace 2D backbone with I3D (ablation) | 3D-RetinaNet · ROAD test, frame-level locations | frame-mAP (%): 30.8 -> 34.9 | Improved |
| increase test-time clip length from 8 to 32 frames (ablation) | 3D-RetinaNet 2D · ROAD test, frame-level agents | frame-mAP (%): 39.5 -> 39.8 | No clear effect |
| increase test-time clip length from 8 to 32 frames (ablation) | 3D-RetinaNet I3D · ROAD test, frame-level agents | frame-mAP (%): 39.5 -> 39.2 | No clear effect |
| increase test-time clip length from 8 to 32 frames (ablation) | 3D-RetinaNet Slowfast · ROAD test, frame-level agents | frame-mAP (%): 47.5 -> 43.7 | Hurt |
| use I3D backbone instead of 2D on UCF-101-24 (ablation) | 3D-RetinaNet · UCF-101-24 video-mAP@0.5 | video-mAP (%): 48.6 -> 58.2 | Improved |
| use I3D backbone instead of 2D on UCF-101-24 (ablation) | 3D-RetinaNet · UCF-101-24 f-mAP | frame-mAP (%): 65.2 -> 75.2 | Improved |
| apply YOLOv5 instead of 3D-RetinaNet for agent detection | YOLOv5 · ROAD test, video-level delta=0.2 agents | video-mAP (%): 29.0 -> 43.3 | Improved |
| apply YOLOv5 instead of 3D-RetinaNet for agent detection | YOLOv5 · ROAD test, frame-level agents | frame-mAP (%): 47.5 -> 56.9 | Improved |
| move from UCF-101-24 to ROAD with the same I3D baseline | 3D-RetinaNet I3D · video-mAP at delta 0.5 | video-mAP (%): 58.2 -> 4.7 | Hurt |
Closest papers by meaning
- One Million Scenes for Autonomous Driving: ONCE Dataset (NeurIPS 2021 · similarity 0.89)
- Ithaca365: Dataset and Driving Perception Under Repeated and Challenging Weather Conditions (CVPR 2022 · similarity 0.88)
- The Road Ahead in Autonomous Driving: The KITScenes Multimodal Dataset (NeurIPS 2026 · similarity 0.86)
- Autonomous Driving Research Requires a Community-Driven Data Paradigm (NeurIPS 2026 · similarity 0.86)
- RoadSceneBench: A Lightweight Benchmark for Mid-Level Road Scene Understanding (CVPR 2026 · similarity 0.86)
- On-Road Object Importance Estimation: A New Dataset and A Model with Multi-Fold Top-Down Guidance (NeurIPS 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.
