Anchor3DLane: Learning to Regress 3D Anchors for Monocular 3D Lane Detection

Benchmark Model Rank Results
3d-lane-detection-on-apollo-synthetic-3d-laneAnchor3DLane#1F1: 95.6X error near: 0.052X error far: 0.306Z error near: 0.015
3d-lane-detection-on-apollo-synthetic-3d-laneAnchor3DLane† (iterative regression)#2F1: 95.4X error near: 0.048X error far: 0.299Z error near: 0.013
3d-lane-detection-on-openlaneAnchor3DLane-T† (multi-frame + iterative regression)#5F1 (all): 54.3Up & Down: 47.2Curve: 58.0Extreme Weather: 52.7
3d-lane-detection-on-openlaneAnchor3DLane† (iterative regression)#6F1 (all): 53.7Up & Down: 46.7Curve: 57.2Extreme Weather: 52.5
3d-lane-detection-on-openlaneAnchor3DLane (ResNet-18)#7F1 (all): 53.1Up & Down: 45.5Curve: 56.2Extreme Weather: 51.9