Determining Vehicle Turn Counts at Multiple Intersections by Separated Vehicle Classes Using CNNs

Jan Folenta, Jakub Spanhel, Vojtech Bartl, Adam Herout; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 596-597

Abstract


In our submission to the NVIDIA AI City Challenge 2020, we address the problem of counting vehicles by their class at multiple intersections. Our solution is based on counting by tracking principle using convolutional neural networks in detection and tracking steps of the proposed method. We have achieved 6th place on the dataset part "A" of Track 1 with score S1 Total = 0.8829, (mwRMSE = 4.3616, S1 Effectiveness = 0.9094, S1 Efficiency = 0.8212). The proposed solution was placed at sixth place in the overall ranking on dataset part A.

Related Material


[pdf]
[bibtex]
@InProceedings{Folenta_2020_CVPR_Workshops,
author = {Folenta, Jan and Spanhel, Jakub and Bartl, Vojtech and Herout, Adam},
title = {Determining Vehicle Turn Counts at Multiple Intersections by Separated Vehicle Classes Using CNNs},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2020}
}