Interactive4D：交互式 4D LiDAR 分割
摘要
交互式分割在促进未来 LiDAR 数据集的标注过程中发挥着重要作用。现有方法依次对每个 LiDAR扫描中的单个对象进行分割，在整个序列中重复该过程，这种做法冗余且不有效。本文提出交互式 4D 分割，这是一种新的范式，允许在多个 LiDAR扫描中同时分割多个对象。我们提出Interactive4D，这是首个能够通过利用 LiDAR数据的序列性质，在单个迭代中对叠加的连续 LiDAR扫描中的多个对象进行分割的交互式 4D 分割模型。在进行交互式分割时，我们的模型利用整个时空体积，从而实现更高效的分割。 operating on the 4D volume, it directly provides consistent instance IDs over time and also simplifies tracking annotations. Moreover, we show that click simulations are crucial for successful model training on LiDAR point clouds. To this end, we design a click simulation strategy that is better suited for the characteristics of LiDAR data. To demonstrate its accuracy and effectiveness, we evaluate Interactive4D on multiple LiDAR datasets, where Interactive4D achieves a new state-of-the-art by a large margin. We publicly release the code and models at https://vision.rwth-aachen.de/Interactive4D.
引用
@article{arxiv.2410.08206,
title = {Interactive4D: Interactive 4D LiDAR Segmentation},
author = {Ilya Fradlin and Idil Esen Zulfikar and Kadir Yilmaz and Theodora Kontogianni and Bastian Leibe},
journal= {arXiv preprint arXiv:2410.08206},
year = {2025}
}
备注
Accepted to ICRA2025!