We investigate the potential of recurrentneural networks (RNNs) to improve traditional on-line multi-target tracking of traffic participants froman ego-vehicle perspective. To this end, we builda modular tracking framework, based on interact-ing multiple models (IMM) and unscented Kalmanfilters (UKF). Following the tracking-by-detectionparadigm, we leverage geometric target propertiesprovided by publicly available 3D object detectors.We then train and integrate two RNNs: A state pre-diction network replaces hand-crafted motion mod-els in our filters and a data association network findsdetection-to-track assignment probabilities. In ourextensive evaluation on the publicly available KITTIdataset we show that our trained models achievecompetitive results and are significantly more robustin the case of unreliable object detections.
|Title of host publication||Proceedings of the 25th Computer Vision Winter Workshop (CVWW)|
|Place of Publication||Ljubljana|
|Publisher||Slovenian Pattern Recognition Society|
|Publication status||Published - 2020|
|Event||25th Computer Vision Winter Workshop: CVWW 2020 - Rogaska Slatina, Slovenia|
Duration: 3 Feb 2020 → 5 Feb 2020
|Conference||25th Computer Vision Winter Workshop|
|Abbreviated title||CVWW 2020|
|Period||3/02/20 → 5/02/20|
Fruhwirth-Reisinger, C., Krispel, G., Possegger, H., & Bischof, H. (2020). Towards Data-driven Multi-target Tracking for Autonomous Driving. In Proceedings of the 25th Computer Vision Winter Workshop (CVWW) (pp. 27-36). Ljubljana: Slovenian Pattern Recognition Society.