English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In object detection with deep neural networks, the box-wise objectness scoretends to be overconfident, sometimes even indicating high confidence inpresence of inaccurate predictions. Hence, the reliability of the predictionand therefore reliable uncertainties are of highest interest. In this work, wepresent a post processing method that for any given neural network providespredictive uncertainty estimates and quality estimates. These estimates arelearned by a post processing model that receives as input a hand-crafted set oftransparent metrics in form of a structured dataset. Therefrom, we learn twotasks for predicted bounding boxes. We discriminate between true positives($\mathit{IoU}\geq0.5$) and false positives ($\mathit{IoU} < 0.5$) which weterm meta classification, and we predict $\mathit{IoU}$ values directly whichwe term meta regression. The probabilities of the meta classification model aimat learning the probabilities of success and failure and therefore provide amodelled predictive uncertainty estimate. On the other hand, meta regressiongives rise to a quality estimate. In numerical experiments, we use the publiclyavailable YOLOv3 network and the Faster-RCNN network and evaluate metaclassification and regression performance on the Kitti, Pascal VOC and COCOdatasets. We demonstrate that our metrics are indeed well correlated with the$\mathit{IoU}$. For meta classification we obtain classification accuracies ofup to 98.92% and AUROCs of up to 99.93%. For meta regression we obtain an $R^2$value of up to 91.78%. These results yield significant improvements compared toother network's objectness score and other baseline approaches. Therefore, weobtain more reliable uncertainty and quality estimates which is particularlyinteresting in the absence of ground truth.

MetaDetect: Uncertainty Quantification and Prediction Quality Estimates for Object Detection