The KITTI Vision Benchmark Suite

Method

monodab [mdab]
[Anonymous Submission]

Submitted on 25 Mar. 2024 14:38 by
[Anonymous Submission]

Running time:		0.02 s
Environment:		1 core @ 2.5 Ghz (Python)

Method Description:

Based on monodetr, change the detr in the image detection framework to dab detr to improve the detection efficiency of the model. After generating the depth detection module of the foreground image, add a depth feature enhancement module to enhance the extracted depth feature information.

Parameters:

type: 'adamw'
lr: 0.0001
weight_decay: 0.0001
mode: LID
num_depth_bins: 80
depth_min: 1e-3
depth_max: 60.

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)	0.00 %	0.00 %	0.00 %
Car (Orientation)	0.00 %	0.00 %	0.00 %
Car (3D Detection)	10.52 %	6.94 %	5.18 %
Car (Bird's Eye View)	17.81 %	11.47 %	9.08 %
Pedestrian (Detection)	0.02 %	0.02 %	0.02 %
Pedestrian (Orientation)	0.02 %	0.02 %	0.02 %
Pedestrian (3D Detection)	1.92 %	1.07 %	1.03 %
Pedestrian (Bird's Eye View)	3.68 %	2.43 %	2.26 %
Cyclist (Detection)	0.00 %	0.00 %	0.00 %
Cyclist (Orientation)	0.00 %	0.00 %	0.00 %
Cyclist (3D Detection)	5.32 %	2.85 %	2.28 %
Cyclist (Bird's Eye View)	6.94 %	3.38 %	3.21 %