Method

Erosion and Cluster with Prior (No ML at all) [ECPrior (No ML)]
https://github.com/larksq/lane_detector_for_KITTI

Submitted on 3 Aug. 2020 18:14 by
Alan Sun (Washington University in St. Louis)

Running time:1 s
Environment:1 core @ 2.5 Ghz (Python)

Method Description:
This method leverage the prior knowledge of where
the lines start and the length of the current lane.
It uses only simple directional erosion and cluster
and can get similar results compared to other Deep
NN methods.
Parameters:
See the parameter file from the github repo.
Latex Bibtex:

Evaluation in Bird's Eye View


Benchmark MaxF AP PRE REC FPR FNR
UM_LANE 87.31 % 74.26 % 87.06 % 87.56 % 2.29 % 12.44 %
This table as LaTeX

Behavior Evaluation


Benchmark PRE-20 F1-20 HR-20 PRE-30 F1-30 HR-30 PRE-40 F1-40 HR-40
UM_LANE 98.70 % 97.95 % 96.22 % 98.65 % 97.28 % 93.96 % 96.70 % 91.86 % 80.00 %
This table as LaTeX

Road/Lane Detection

The following plots show precision/recall curves for the bird's eye view evaluation.


Distance-dependent Behavior Evaluation

The following plots show the F1 score/Precision/Hitrate with respect to the longitudinal distance which has been used for evaluation.



This figure as: png eps pdf

This figure as: png eps pdf

This figure as: png eps pdf

Visualization of Results

The following images illustrate the performance of the method qualitatively on a couple of test images. We first show results in the perspective image, followed by evaluation in bird's eye view. Here, red denotes false negatives, blue areas correspond to false positives and green represents true positives.



This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png

This figure as: png


eXTReMe Tracker