Reliable epistemic uncertainty estimation is an essential component for backend applications of deep object detectors in safety-critical environments. Modern network architectures tend to give poorly calibrated confidences with limited predictive power. Here, we introduce novel gradient-based uncertainty metrics and investigate them for different object detection architectures. Experiments on the MS COCO, PASCAL VOC and the KITTI dataset show significant improvements in true positive / false positive discrimination and prediction of intersection over union as compared to network confidence. We also find improvement over Monte-Carlo dropout uncertainty metrics and further significant boosts by aggregating different sources of uncertainty metrics. The resulting uncertainty models generate well-calibrated confidences in all instances. Furthermore, we implement our uncertainty quantification models into object detection pipelines as a means to discern true against false predictions, replacing the ordinary score-threshold-based decision rule. In our experiments, we achieve a significant boost in detection performance in terms of mean average precision. With respect to computational complexity, we find that computing gradient uncertainty metrics results in floating point operation counts similar to those of Monte-Carlo dropout.

No abstract

No abstract

Safety-critical applications of deep neural networks require reliable confidence estimation methods with high predictive power. However, evaluating and comparing different methods for uncertainty quantification is oftentimes highly context-dependent. In this chapter, we introduce flexible evaluation protocols which are applicable to a wide range of tasks with an emphasis on object detection. In this light, we investigate uncertainty metrics based on the network output, as well as metrics based on a learning gradient, both of which significantly outperform the confidence score of the network. While output-based uncertainty is produced by post-processing steps and is computationally efficient, gradient-based uncertainty, in principle, allows for localization of uncertainty within the network architecture. We show that both sources of uncertainty are mutually non-redundant and can be combined beneficially. Furthermore, we show direct applications of uncertainty quantification by improving detection accuracy.

Currently, in routine diagnostics, most molecular testing is performed on formalin-fixed, paraffin-embedded tissue after a histomorphological assessment. In order to find the best possible and targeted individual therapy, knowing the mutational status of the tumour is crucial. The “AVENIO Millisect” system Roche introduced an automation solution for the dissection of tissue on slides. This technology allows the precise and fully automated dissection of the tumour area without wasting limited and valuable patient material. In this study, the digitally guided microdissection was directly compared to the manual macrodissection regarding the precision and duration of the procedure, their DNA concentrations as well as DNA qualities, and the overall costs in 24 FFPE samples. In 21 of 24 cases (87.5%), the DNA yields of the manually dissected samples were higher in comparison to the automatically dissected samples. Shorter execution times and lower costs were also benefits of the manual scraping process. Nevertheless, the DNA quality achieved with both methods was comparable, which is essential for further molecular testing. Therefore, it could be used as an additional tool for precise tumour enrichment.