Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning

Jisu Kang; Wooseok Shin; Jin Sob Kim; Hyun Joon Park; Yujin Ham; Sung Won Han

doi:10.1609/aaai.v39i28.35264

Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning

Jisu Kang
, Wooseok Shin
, Jin Sob Kim
, Hyun Joon Park
, Yujin Ham
, Sung Won Han^*

^*Corresponding author for this work

School of Industrial Management Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

The task of 3D object detection is crucial for various applications that rely on identifying objects in three-dimensional space using inputs like LiDAR point clouds and images. However, LiDAR-based detection faces challenges due to the sparsity of point clouds, especially at greater distances. To address this, depth completion models have been used to generate virtual points from RGB images, but they struggle with real-time applications due to high computational costs. Our work eliminates the depth completion process, significantly improving processing speed while minimizing performance degradation. Consequently, our method has achieved an optimal balance between speed and accuracy on the KITTI leaderboard.

Original language	English
Title of host publication	Special Track on AI Alignment
Editors	Toby Walsh, Julie Shah, Zico Kolter
Publisher	Association for the Advancement of Artificial Intelligence
Pages	29396-29398
Number of pages	3
Edition	28
ISBN (Electronic)	157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 157735897X, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978, 9781577358978
DOIs	https://doi.org/10.1609/aaai.v39i28.35264
Publication status	Published - 2025 Apr 11
Event	39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025 - Philadelphia, United States Duration: 2025 Feb 25 → 2025 Mar 4

Publication series

Name	Proceedings of the AAAI Conference on Artificial Intelligence
Number	28
Volume	39
ISSN (Print)	2159-5399
ISSN (Electronic)	2374-3468

Conference

Conference	39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025
Country/Territory	United States
City	Philadelphia
Period	25/2/25 → 25/3/4

Bibliographical note

Publisher Copyright:
Copyright © 2025, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.

ASJC Scopus subject areas

Artificial Intelligence

Access to Document

10.1609/aaai.v39i28.35264

Cite this

Kang, J., Shin, W., Kim, J. S., Park, H. J., Ham, Y., & Han, S. W. (2025). Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning. In T. Walsh, J. Shah, & Z. Kolter (Eds.), Special Track on AI Alignment (28 ed., pp. 29396-29398). (Proceedings of the AAAI Conference on Artificial Intelligence; Vol. 39, No. 28). Association for the Advancement of Artificial Intelligence. https://doi.org/10.1609/aaai.v39i28.35264

Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning. / Kang, Jisu; Shin, Wooseok; Kim, Jin Sob et al.
Special Track on AI Alignment. ed. / Toby Walsh; Julie Shah; Zico Kolter. 28. ed. Association for the Advancement of Artificial Intelligence, 2025. p. 29396-29398 (Proceedings of the AAAI Conference on Artificial Intelligence; Vol. 39, No. 28).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Kang, J, Shin, W, Kim, JS, Park, HJ, Ham, Y & Han, SW 2025, Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning. in T Walsh, J Shah & Z Kolter (eds), Special Track on AI Alignment. 28 edn, Proceedings of the AAAI Conference on Artificial Intelligence, no. 28, vol. 39, Association for the Advancement of Artificial Intelligence, pp. 29396-29398, 39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025, Philadelphia, United States, 25/2/25. https://doi.org/10.1609/aaai.v39i28.35264

Kang J, Shin W, Kim JS, Park HJ, Ham Y, Han SW. Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning. In Walsh T, Shah J, Kolter Z, editors, Special Track on AI Alignment. 28 ed. Association for the Advancement of Artificial Intelligence. 2025. p. 29396-29398. (Proceedings of the AAAI Conference on Artificial Intelligence; 28). doi: 10.1609/aaai.v39i28.35264

Kang, Jisu ; Shin, Wooseok ; Kim, Jin Sob et al. / Beyond Virtual Points : Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning. Special Track on AI Alignment. editor / Toby Walsh ; Julie Shah ; Zico Kolter. 28. ed. Association for the Advancement of Artificial Intelligence, 2025. pp. 29396-29398 (Proceedings of the AAAI Conference on Artificial Intelligence; 28).

@inproceedings{a94d2843fda944c48b9d6fd7806fdf99,

title = "Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning",

abstract = "The task of 3D object detection is crucial for various applications that rely on identifying objects in three-dimensional space using inputs like LiDAR point clouds and images. However, LiDAR-based detection faces challenges due to the sparsity of point clouds, especially at greater distances. To address this, depth completion models have been used to generate virtual points from RGB images, but they struggle with real-time applications due to high computational costs. Our work eliminates the depth completion process, significantly improving processing speed while minimizing performance degradation. Consequently, our method has achieved an optimal balance between speed and accuracy on the KITTI leaderboard.",

author = "Jisu Kang and Wooseok Shin and Kim, \{Jin Sob\} and Park, \{Hyun Joon\} and Yujin Ham and Han, \{Sung Won\}",

note = "Publisher Copyright: Copyright {\textcopyright} 2025, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.; 39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025 ; Conference date: 25-02-2025 Through 04-03-2025",

year = "2025",

month = apr,

day = "11",

doi = "10.1609/aaai.v39i28.35264",

language = "English",

series = "Proceedings of the AAAI Conference on Artificial Intelligence",

publisher = "Association for the Advancement of Artificial Intelligence",

number = "28",

pages = "29396--29398",

editor = "Toby Walsh and Julie Shah and Zico Kolter",

booktitle = "Special Track on AI Alignment",

edition = "28",

}

TY - GEN

T1 - Beyond Virtual Points

T2 - 39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

AU - Kang, Jisu

AU - Shin, Wooseok

AU - Kim, Jin Sob

AU - Park, Hyun Joon

AU - Ham, Yujin

AU - Han, Sung Won

PY - 2025/4/11

Y1 - 2025/4/11

N2 - The task of 3D object detection is crucial for various applications that rely on identifying objects in three-dimensional space using inputs like LiDAR point clouds and images. However, LiDAR-based detection faces challenges due to the sparsity of point clouds, especially at greater distances. To address this, depth completion models have been used to generate virtual points from RGB images, but they struggle with real-time applications due to high computational costs. Our work eliminates the depth completion process, significantly improving processing speed while minimizing performance degradation. Consequently, our method has achieved an optimal balance between speed and accuracy on the KITTI leaderboard.

AB - The task of 3D object detection is crucial for various applications that rely on identifying objects in three-dimensional space using inputs like LiDAR point clouds and images. However, LiDAR-based detection faces challenges due to the sparsity of point clouds, especially at greater distances. To address this, depth completion models have been used to generate virtual points from RGB images, but they struggle with real-time applications due to high computational costs. Our work eliminates the depth completion process, significantly improving processing speed while minimizing performance degradation. Consequently, our method has achieved an optimal balance between speed and accuracy on the KITTI leaderboard.

UR - https://www.scopus.com/pages/publications/105003968809

U2 - 10.1609/aaai.v39i28.35264

DO - 10.1609/aaai.v39i28.35264

M3 - Conference contribution

AN - SCOPUS:105003968809

T3 - Proceedings of the AAAI Conference on Artificial Intelligence

SP - 29396

EP - 29398

BT - Special Track on AI Alignment

A2 - Walsh, Toby

A2 - Shah, Julie

A2 - Kolter, Zico

PB - Association for the Advancement of Artificial Intelligence

Y2 - 25 February 2025 through 4 March 2025

ER -

Beyond Virtual Points: Depth-Enhanced LiDAR-only 3D Object Detection with Semi-Supervised Learning

Abstract

Publication series

Conference

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this