The KITTI Vision Benchmark Suite

Method

Cube R-CNN [Cube R-CNN]
https://github.com/facebookresearch/omni3d

Submitted on 27 Mar. 2023 08:16 by
Garrick Brazil (Michigan State University)

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

Method Description:

A single image baseline method trained on KITTI's
full set with 7,481 images and 5 categories.

Parameters:

We use scale augmentation and virtual focal length
of 512 to train Cube R-CNN for 128 epochs.

Latex Bibtex:

@inproceedings{brazil2023omni3d,
author = {Garrick Brazil and Abhinav Kumar
and Julian Straub and Nikhila Ravi and Justin
Johnson and Georgia Gkioxari},
title = {{Omni3D}: A Large Benchmark and
Model for {3D} Object Detection in the Wild},
booktitle = {CVPR},
address = {Vancouver, Canada},
month = {June},
year = {2023},
organization = {IEEE},
}

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)	95.78 %	92.72 %	84.81 %
Car (Orientation)	38.09 %	39.78 %	35.40 %
Car (3D Detection)	23.59 %	15.01 %	12.56 %
Car (Bird's Eye View)	31.70 %	21.20 %	18.43 %
Pedestrian (Detection)	64.96 %	53.27 %	47.84 %
Pedestrian (Orientation)	34.26 %	28.07 %	25.14 %
Pedestrian (3D Detection)	11.17 %	6.95 %	5.87 %
Pedestrian (Bird's Eye View)	11.67 %	7.65 %	6.60 %
Cyclist (Detection)	81.99 %	66.98 %	58.56 %
Cyclist (Orientation)	29.00 %	23.98 %	21.67 %
Cyclist (3D Detection)	3.65 %	2.67 %	2.28 %
Cyclist (Bird's Eye View)	5.01 %	3.35 %	3.23 %