The object detection task consists of localizing and labeling objects in a given image. Convolutional neural networks (CNNs) have shown remarkable results in computer vision tasks and are the basis for the current state-of-the-art methods. However, CNNs are often overconfident about their predictions. In the paper "Assessing different box merging strategies and uncertainty estimation methods in multimodel object detection" Felippe Schmoeller da Roza and his colleagues at Fraunhofer IKS compare different methods that provide uncertainty estimates for object detectors, allowing such models to express uncertainty levels for the position of detected objects.
The paper will be presented during the workshop »Beyond mAP: Reassessing the Evaluation of Object Detectors« that is part of the European Conference on Computer Vision (ECCV2020).
Talk at the ECCV2020 workshop "Beyond mAP: Reassessing the Evaluation of Object Detectors"
Felippe Schmoeller da Roza
"Assessing different box merging strategies and uncertainty estimation methods in multimodel object detection"
August 28, 2020, 09:00 a.m. CET