How to download MORDA
The link to sign up and request the dataset access will be available in future.
Deep-neural-network (DNN) based perception models are indispensable in the development of autonomous vehicles (AVs). However, their reliance on large-scale, high-quality data is broadly recognized as a burdensome necessity due to the substantial cost of data acquisition and labeling.
Further, the issue is not a one-time concern, as AVs might need a new dataset if they are to be deployed to another region (Real-target domain) that the in-hand dataset within real-source domain cannot incorporate. To mitigate this burden, this paper aims to showcase the efficacy of synthetic environments as an auxiliary domain where indirect experience of unseen real- world regions is available in a time- and cost-effective manner. In particular, we reconstruct the digital twins of the real-target domain (South Korea, DRealTrg) and data-acquisition framework of the real-source domain (nuScenes, DRealSrc) within a simulator. Next, exploiting the synthetic world (DSynSrc+Trg), we generate our novel dataset, MORDA: Mixture Of Real-domain character- istics for synthetic-data-assisted Domain Adaptation. As MORDA provides multi-view images and point cloud data along with rich bounding-box labels, we conducted comprehensive experiments with 2D/3D object detectors. To verify the value of MORDA regarding the transferability of reproduced synthetic features, 2D/3D detectors are trained solely on the combined training set of nuScenes and MORDA. Afterward, their performance is evaluated in unforeseen DRealTrg1. Our extensive experiments present the significant gain of mean Average Precision (mAP) for DRealTrg while their performance in DRealSrc is retained or even enhanced.
The link to sign up and request the dataset access will be available in future.
Although we have taken careful steps to remove or obscure personally identifiable information (PII), some content requiring further removal may still remain in the dataset. We kindly ask you to submit a modification request using this link if you observe any content that needs additional review or removal.
@article{lim2024morda,
author = {Lim, Hojun and Yoo, Heecheol and Lee, Jinwoo and Jeon, Seungmin and Jeon, Hyeongseok},
title = {TBD},
journal = {TBD},
year = {TBD},
}