Method

FPN_Net [FNETMS]
[Anonymous Submission]

Submitted on 8 Oct. 2017 04:30 by
[Anonymous Submission]

 Running time: 0.04 s Environment: GPU @ 2.0 Ghz (Python + C/C++)

 Method Description: using features from layer to layer and fusion with IPM Parameters: \alpha=0.2 Latex Bibtex: \no

Evaluation in Bird's Eye View

 Benchmark MaxF AP PRE REC FPR FNR UM_ROAD 94.51 % 92.72 % 94.92 % 94.11 % 2.30 % 5.89 % UMM_ROAD 96.08 % 94.95 % 95.56 % 96.60 % 4.93 % 3.40 % UU_ROAD 93.71 % 91.79 % 94.08 % 93.34 % 1.91 % 6.66 % URBAN_ROAD 94.99 % 93.18 % 94.90 % 95.09 % 2.82 % 4.91 %
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