Method

Real-Time Road Segmentation Using LiDAR Data Processing on an FPGA [la] [LiDAR-SPHnet]
[Anonymous Submission]

Submitted on 24 Oct. 2017 14:01 by
[Anonymous Submission]

Running time:52 ms
Environment:FPGA @ 200 Mhz

Method Description:
pre-processing: LiDAR data sampled into Spherical
grid.
Neural Network: Non Scale 5x5 filter stacks
post-processing: Polygon filling up
Parameters:
Training Solver: Adam
lr=1e-6
Training epoch:60
Latex Bibtex:

Evaluation in Bird's Eye View


Benchmark MaxF AP PRE REC FPR FNR
UM_ROAD 90.85 % 86.27 % 88.81 % 92.98 % 5.34 % 7.02 %
UMM_ROAD 91.40 % 91.89 % 89.03 % 93.90 % 12.72 % 6.10 %
UU_ROAD 86.36 % 84.25 % 85.33 % 87.41 % 4.90 % 12.59 %
URBAN_ROAD 89.60 % 87.99 % 86.35 % 93.12 % 8.11 % 6.88 %
This table as LaTeX

Behavior Evaluation


This table as LaTeX

Road/Lane Detection

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



This figure as: png eps pdf

This figure as: png eps pdf

This figure as: png eps pdf

This figure as: png eps pdf

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.


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

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