Uncertainty Estimation in One-Stage Object Detection

Florian Kraus1, Klaus Dietmayer2

  • 1Daimler AG
  • 2University of Ulm



Regular Session


11:00 - 12:00 | Mon 28 Oct | The Great Room II | MoC-T3

Regular Session on Object Detection and Classification (I)

Full Text


Environment perception is the task in autonomous driving on which all subsequent steps rely. A key part of perception is to safely detect other road users such as vehicles, pedestrians,andcyclists.Withmoderndeeplearningtechniques huge progress was made in the last years. However such deep learning based object detection models cannot predict how certain they are in their predictions, potentially hampering the performance of later steps such as tracking or sensor fusion. We present a viable approaches to estimate uncertainty in an one-stage object detector, while improving the detection performance of the baseline approach. The proposed model is evaluated on a large scale automotive pedestrian dataset. Experimental results show that the uncertainty is coupled with detection accuracy and occlusion of pedestrians.

Additional Information

No information added


No videos found