Method

Normalized Pixel Position Extend Network [Pos-ex]
http://ethereon.github.io/netscope/#/gist/5279e4243d0b4fa553f229ac257f5bfb

Submitted on 16 Jul. 2017 20:24 by
Yecheng Lyu (Worcester Polytechnic Institute)

Running time:120 ms
Environment:GPU(K20) @ 0.7 Ghz (Matlab + Caffe)

Method Description:
(3x3x24+5x5x24)x4 Conv Layer with Normalized
Position Layer
Parameters:
Adam 1e-5 for 15 epochs
Latex Bibtex:

Evaluation in Bird's Eye View


Benchmark MaxF AP PRE REC FPR FNR
UM_ROAD 79.47 % 85.99 % 78.78 % 80.17 % 9.84 % 19.83 %
UMM_ROAD 87.75 % 90.75 % 82.54 % 93.67 % 21.78 % 6.33 %
UU_ROAD 74.60 % 79.21 % 70.52 % 79.19 % 10.79 % 20.81 %
URBAN_ROAD 81.34 % 86.25 % 77.30 % 85.81 % 13.88 % 14.19 %
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