Method

Using depth information to estimate the scale [CNN-ds]
[Anonymous Submission]

Submitted on 21 May. 2018 12:03 by
[Anonymous Submission]

Running time:0.05 s
Environment:1 core @ 2.5 Ghz (Python)

Method Description:
Using depth to calcuate the scale of objects
instead of using lot of anchors
Parameters:
scale=2
Latex Bibtex:
no

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) 86.86 % 87.15 % 70.77 %
Car (Orientation) 46.39 % 48.12 % 38.40 %
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




eXTReMe Tracker