Method

4d-MSCNN [st] [4d-MSCNN]
https://github.com/ferrazpedro1/4d-mscnn

Submitted on 4 Oct. 2020 21:45 by
Pedro Augusto Pinho Ferraz (Pontifical Catholic University of Minas Gerais - PUC Minas)

Running time:0.3 min
Environment:GPU @ 3.0 Ghz (Matlab + C/C++)

Method Description:
Three-stage RGBD architecture for vehicle and pedestrian detection using convolutional neural networks and stereo vision (re-submission from 2018)

The method uses the depth information extracted from the stereo camera setup to add a 4th layer into the RGB image (creating an RGBD input) and uses it as input to the detection CNN.
Parameters:
Same training hyperparameters from the original MSCNN.
Latex Bibtex:
@article{ferraz2020three,
title={Three-stage RGBD architecture for vehicle and pedestrian detection using convolutional neural networks and stereo vision},
author={Ferraz, Pedro Augusto Pinho and de Oliveira, Bernardo Augusto Godinho and Ferreira, Fl{\'a}via Magalh{\~a}es Freitas Ferreira and da Silva Martins, Carlos Augusto Paiva and others},
journal={IET Intelligent Transport Systems},
year={2020},
publisher={IET}
}

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) 92.40 % 89.37 % 77.00 %
This table as LaTeX


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




eXTReMe Tracker