Realistically distributing object placements in synthetic training data improves the performance of vision-based object detection models

Setareh Dabiri; Vasileios Lioutas; Berend Zwartsenberg; Yunpeng Liu; Matthew Niedoba; Xiaoxuan Liang; Dylan Green; Justice Sefas; Jonathan Wilder Lavington; Frank Wood; Adam Scibior

Realistically distributing object placements in synthetic training data improves the performance of vision-based object detection models

Computer Vision and Pattern Recognition 2023-05-25 v1

Authors: Setareh Dabiri , Vasileios Lioutas , Berend Zwartsenberg , Yunpeng Liu , Matthew Niedoba , Xiaoxuan Liang , Dylan Green , Justice Sefas , Jonathan Wilder Lavington , Frank Wood , Adam Scibior

View on arXiv ↗ PDF ↗

Abstract

When training object detection models on synthetic data, it is important to make the distribution of synthetic data as close as possible to the distribution of real data. We investigate specifically the impact of object placement distribution, keeping all other aspects of synthetic data fixed. Our experiment, training a 3D vehicle detection model in CARLA and testing on KITTI, demonstrates a substantial improvement resulting from improving the object placement distribution.

Keywords

synthetic data generation object detection

Cite

@article{arxiv.2305.14621,
  title  = {Realistically distributing object placements in synthetic training data improves the performance of vision-based object detection models},
  author = {Setareh Dabiri and Vasileios Lioutas and Berend Zwartsenberg and Yunpeng Liu and Matthew Niedoba and Xiaoxuan Liang and Dylan Green and Justice Sefas and Jonathan Wilder Lavington and Frank Wood and Adam Scibior},
  journal= {arXiv preprint arXiv:2305.14621},
  year   = {2023}
}

Realistically distributing object placements in synthetic training data improves the performance of vision-based object detection models

Abstract

Keywords

Cite

Related papers