Method

VGG Finetuned FCN [VGGFCN]
[Anonymous Submission]

Submitted on 19 Feb. 2016 10:42 by
[Anonymous Submission]

Running time:0.4 s
Environment:GPU @ 1.0 Ghz (Python + C/C++)

Method Description:
16 Layer VGG Imagenet trained network fully
convolutionalized and finetuned on Kitti.
Parameters:
lr=e^-10
Latex Bibtex:

Evaluation in Bird's Eye View


Benchmark MaxF AP PRE REC FPR FNR
UM_ROAD 90.93 % 83.56 % 88.79 % 93.18 % 5.36 % 6.82 %
UMM_ROAD 94.26 % 91.14 % 95.02 % 93.51 % 5.38 % 6.49 %
UU_ROAD 88.90 % 76.87 % 88.49 % 89.31 % 3.79 % 10.69 %
URBAN_ROAD 91.95 % 86.42 % 91.51 % 92.40 % 4.72 % 7.61 %
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.


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