Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Bian, Wenjing, Wang, Zirui, Vedaldi, Andrea
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2408.12747
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913477246844928
author Bian, Wenjing
Wang, Zirui
Vedaldi, Andrea
author_facet Bian, Wenjing
Wang, Zirui
Vedaldi, Andrea
contents Image-based 3D object detection is widely employed in applications such as autonomous vehicles and robotics, yet current systems struggle with generalisation due to complex problem setup and limited training data. We introduce a novel pipeline that decouples 3D detection from 2D detection and depth prediction, using a diffusion-based approach to improve accuracy and support category-agnostic detection. Additionally, we introduce the Normalised Hungarian Distance (NHD) metric for an accurate evaluation of 3D detection results, addressing the limitations of traditional IoU and GIoU metrics. Experimental results demonstrate that our method achieves state-of-the-art accuracy and strong generalisation across various object categories and datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2408_12747
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle CatFree3D: Category-agnostic 3D Object Detection with Diffusion
Bian, Wenjing
Wang, Zirui
Vedaldi, Andrea
Computer Vision and Pattern Recognition
Image-based 3D object detection is widely employed in applications such as autonomous vehicles and robotics, yet current systems struggle with generalisation due to complex problem setup and limited training data. We introduce a novel pipeline that decouples 3D detection from 2D detection and depth prediction, using a diffusion-based approach to improve accuracy and support category-agnostic detection. Additionally, we introduce the Normalised Hungarian Distance (NHD) metric for an accurate evaluation of 3D detection results, addressing the limitations of traditional IoU and GIoU metrics. Experimental results demonstrate that our method achieves state-of-the-art accuracy and strong generalisation across various object categories and datasets.
title CatFree3D: Category-agnostic 3D Object Detection with Diffusion
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2408.12747