Exploring Active 3D Object Detection from a Generalization Perspective

Authors: Yadan Luo, Zhuoxiao Chen, Zijian Wang, Xin Yu, Zi Huang, Mahsa Baktashmotlagh

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To validate the effectiveness and applicability of CRB, we conduct extensive experiments on the two benchmark 3D object detection datasets of KITTI and Waymo and examine both one-stage (i.e., SECOND) and two-stage 3D detectors (i.e., PVRCNN).
Researcher Affiliation Academia Yadan Luo , Zhuoxiao Chen , Zijian Wang, Xin Yu, Zi Huang, Mahsa Baktashmotlagh The University of Queensland, Australia
Pseudocode Yes The algorithm is summarized in the supplemental material.
Open Source Code Yes Source code: https://github.com/Luoyadan/CRB-active-3Ddet.
Open Datasets Yes Datasets. KITTI (Geiger et al., 2012) is one of the most representative datasets for point cloud based object detection. The Waymo Open dataset (Sun et al., 2020) is a challenging testbed for autonomous driving, containing 158,361 training samples and 40,077 testing samples.
Dataset Splits Yes The dataset consists of 3,712 training samples (i.e., point clouds) and 3,769 val samples.
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU types, or memory specifications used for running experiments.
Software Dependencies No The paper mentions developing an 'active-3D-det toolbox' but does not specify any software dependencies with version numbers (e.g., Python version, PyTorch version, CUDA version).
Experiment Setup Yes The K1 and K2 are empirically set to 300 and 200 for KITTI and 2, 000 and 1, 200 for Waymo. We specify the settings of hyper-parameters, the training scheme and the implementation details of our model and AL baselines in Sec. B of the supplementary material.