Research map / Video / Video segmentation and tracking
Homography Guided Temporal Fusion for Road Line and Marking Segmentation
Shan Wang, Chuong Nguyen, Jiawei Liu, Kaihao Zhang, Wenhan Luo, Yanhao Zhang, Sundaram Muthu, Fahira Afzal Maken, Hongdong Li
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 |
|---|---|---|---|
| add HomoFusion temporal fusion (4 frames vs 1 frame) (ablation) | HomoFusion · ApolloScape (18 mIoU) | mIoU: 49.4 -> 59.3 | Improved |
| add homography guidance instead of identity correspondence (ablation) | HomoFusion · ApolloScape (18 mIoU) | mIoU: 51.8 -> 59.3 | Improved |
| add RSNE road normal estimation (ablation) | HomoFusion · ApolloScape (18 mIoU) | mIoU: 57.4 -> 59.3 | Improved |
| estimate normal with deep features rather than RGB (ablation) | RSNE · ApolloScape (18 mIoU) | mIoU: 54.9 -> 59.3 | Improved |
| add random extrinsic noise (<1 m, up to 30 deg) (ablation) | HomoFusion · ApolloScape (18 mIoU) | mIoU: 59.3 -> 57.5 | Hurt |
| use COLMAP-estimated extrinsics instead of dataset extrinsics (ablation) | HomoFusion · ApolloScape (18 mIoU) | mIoU: 59.3 -> 56.2 | Hurt |
| use OpenCV FindHomography instead of RSNE-guided homography (ablation) | HomoFusion · ApolloScape (18 mIoU) | mIoU: 59.3 -> 47.9 | Hurt |
| increase number of frames beyond 4 | HomoFusion · ApolloScape | mIoU | No clear effect |
| frame gap of 2 versus other gaps | HomoFusion · ApolloScape | mIoU | Improved |
| use lightweight EfficientNet-B6 backbone vs MiT-B1 (ablation) | HomoFusion · ApolloScape (18 mIoU) | mIoU: 57.2 -> 59.3 | Improved |
| apply HomoFusion to water puddle segmentation | HomoFusion · Puddle-1000 On-road | F1-measure: 0.70 -> 0.80 | Improved |
| use HomoFusion instead of global-attention temporal fusion (CFFM) | video segmentation baseline · ApolloScape (18 mIoU) | mIoU: 53.2 -> 59.3 | Improved |
Closest papers by meaning
- Sketch and Refine: Towards Fast and Accurate Lane Detection (AAAI 2024 · similarity 0.85)
- LaneCPP: Continuous 3D Lane Detection using Physical Priors (CVPR 2024 · similarity 0.85)
- LaneSegNet: Map Learning with Lane Segment Perception for Autonomous Driving (ICLR 2024 · similarity 0.85)
- Rethinking Lanes and Points in Complex Scenarios for Monocular 3D Lane Detection (CVPR 2025 · similarity 0.85)
- Controllable Road Marking Generation (NeurIPS 2026 · similarity 0.84)
- Patched Line Segment Learning for Vector Road Mapping (AAAI 2024 · similarity 0.84)
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.
