Method

Deep Convex NMF in Deep Convolutional Networks for On-Road Obstacle Detection [NMF-CNN]
[Anonymous Submission]

Submitted on 16 May. 2016 05:48 by
[Anonymous Submission]

Running time:0.1 s
Environment:GPU @ 2.5 Ghz (Matlab + C/C++)

Method Description:
Our detection system is based on the Faster RCNN
and combining a proposed model, called deep
Convex-NMF.
The overall system is called Deep Convex-NMF CNN.
Parameters:
We are using the 10 scales and 7 aspect ratio.
The dropout ratio is 0.25 instead of 0.5.
Latex Bibtex:

Detailed Results

Object detection and orientation estimation results. Results for object detection are given in terms of average precision (AP) and results for joint object detection and orientation estimation are provided in terms of average orientation similarity (AOS).


Benchmark Easy Moderate Hard
Car (Detection) 77.32 % 62.21 % 49.29 %
Car (Orientation) 32.01 % 26.11 % 19.11 %
Pedestrian (Detection) 65.16 % 49.26 % 45.38 %
Pedestrian (Orientation) 40.14 % 30.94 % 28.58 %
Cyclist (Detection) 56.30 % 42.13 % 37.46 %
Cyclist (Orientation) 22.03 % 16.78 % 15.10 %
This table as LaTeX


2D object detection results.
This figure as: png eps pdf txt gnuplot



Orientation estimation results.
This figure as: png eps pdf txt gnuplot



2D object detection results.
This figure as: png eps pdf txt gnuplot



Orientation estimation results.
This figure as: png eps pdf txt gnuplot



2D object detection results.
This figure as: png eps pdf txt gnuplot



Orientation estimation results.
This figure as: png eps pdf txt gnuplot




eXTReMe Tracker