Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

Shift R-CNN: Deep Monocular 3D Object Detection with Closed-Form Geometric Constraints

Abstract: 

We propose Shift R-CNN, a hybrid model for monocular 3D object detection, which combines deep learning with the power of geometry. We adapt a Faster R-CNN network for regressing initial 2D and 3D object properties and combine it with a least squares solution for the inverse 2D to 3D geometric mapping problem, using the camera projection matrix. The closed-form solution of the mathematical system, along with the initial output of the adapted Faster R-CNN are then passed through a final ShiftNet network that refines the result using our newly proposed Volume Displacement Loss. Our novel, geometrically constrained deep learning approach to monocular 3D object detection obtains top results on KITTI 3D Object Detection Benchmark, being the best among all monocular methods that do not use any pre-trained network for depth estimation.

up
0 users have voted:

Paper Details

Authors:
Andretti Naiden,Vlad Paunescu,Gyeongmo Kim,ByeongMoon Jeon,Marius Leordeanu
Submitted On:
22 September 2019 - 3:45am
Short Link:
Type:
Presentation Slides
Event:
Presenter's Name:
PAUNESCU VLAD,ANDRETTI NAIDEN
Paper Code:
3316
Document Year:
2019
Cite

Document Files

shift-rcnn-icip2019_prefinal.pdf

(11)

Subscribe

[1] Andretti Naiden,Vlad Paunescu,Gyeongmo Kim,ByeongMoon Jeon,Marius Leordeanu, "Shift R-CNN: Deep Monocular 3D Object Detection with Closed-Form Geometric Constraints", IEEE SigPort, 2019. [Online]. Available: http://sigport.org/4815. Accessed: Oct. 18, 2019.
@article{4815-19,
url = {http://sigport.org/4815},
author = {Andretti Naiden;Vlad Paunescu;Gyeongmo Kim;ByeongMoon Jeon;Marius Leordeanu },
publisher = {IEEE SigPort},
title = {Shift R-CNN: Deep Monocular 3D Object Detection with Closed-Form Geometric Constraints},
year = {2019} }
TY - EJOUR
T1 - Shift R-CNN: Deep Monocular 3D Object Detection with Closed-Form Geometric Constraints
AU - Andretti Naiden;Vlad Paunescu;Gyeongmo Kim;ByeongMoon Jeon;Marius Leordeanu
PY - 2019
PB - IEEE SigPort
UR - http://sigport.org/4815
ER -
Andretti Naiden,Vlad Paunescu,Gyeongmo Kim,ByeongMoon Jeon,Marius Leordeanu. (2019). Shift R-CNN: Deep Monocular 3D Object Detection with Closed-Form Geometric Constraints. IEEE SigPort. http://sigport.org/4815
Andretti Naiden,Vlad Paunescu,Gyeongmo Kim,ByeongMoon Jeon,Marius Leordeanu, 2019. Shift R-CNN: Deep Monocular 3D Object Detection with Closed-Form Geometric Constraints. Available at: http://sigport.org/4815.
Andretti Naiden,Vlad Paunescu,Gyeongmo Kim,ByeongMoon Jeon,Marius Leordeanu. (2019). "Shift R-CNN: Deep Monocular 3D Object Detection with Closed-Form Geometric Constraints." Web.
1. Andretti Naiden,Vlad Paunescu,Gyeongmo Kim,ByeongMoon Jeon,Marius Leordeanu. Shift R-CNN: Deep Monocular 3D Object Detection with Closed-Form Geometric Constraints [Internet]. IEEE SigPort; 2019. Available from : http://sigport.org/4815