Uncertainty for Identifying Open-Set Errors in Visual Object Detection
This addresses safety-critical applications in robotics and computer vision by reducing open-set errors, though it is incremental as it builds on existing detector architectures.
The paper tackles the problem of open-set errors in object detection by proposing GMM-Det, a method that uses epistemic uncertainty to identify and reject false positives from unseen classes, showing it outperforms existing techniques with minimal computational overhead.
Deployed into an open world, object detectors are prone to open-set errors, false positive detections of object classes not present in the training dataset. We propose GMM-Det, a real-time method for extracting epistemic uncertainty from object detectors to identify and reject open-set errors. GMM-Det trains the detector to produce a structured logit space that is modelled with class-specific Gaussian Mixture Models. At test time, open-set errors are identified by their low log-probability under all Gaussian Mixture Models. We test two common detector architectures, Faster R-CNN and RetinaNet, across three varied datasets spanning robotics and computer vision. Our results show that GMM-Det consistently outperforms existing uncertainty techniques for identifying and rejecting open-set detections, especially at the low-error-rate operating point required for safety-critical applications. GMM-Det maintains object detection performance, and introduces only minimal computational overhead. We also introduce a methodology for converting existing object detection datasets into specific open-set datasets to evaluate open-set performance in object detection.