ObjectLab: Automated Diagnosis of Mislabeled Images in Object Detection Data
This addresses the issue of brittle object detection systems, such as those in autonomous vehicles, by enabling automated error detection for label review, though it is incremental as it builds on existing models without changing modeling code.
The paper tackles the problem of annotation errors in object detection datasets by proposing ObjectLab, an algorithm that automatically detects mislabeled images, resulting in improved precision and recall compared to existing methods across datasets like COCO and models such as Detectron-X101 and Faster-RCNN.
Despite powering sensitive systems like autonomous vehicles, object detection remains fairly brittle in part due to annotation errors that plague most real-world training datasets. We propose ObjectLab, a straightforward algorithm to detect diverse errors in object detection labels, including: overlooked bounding boxes, badly located boxes, and incorrect class label assignments. ObjectLab utilizes any trained object detection model to score the label quality of each image, such that mislabeled images can be automatically prioritized for label review/correction. Properly handling erroneous data enables training a better version of the same object detection model, without any change in existing modeling code. Across different object detection datasets (including COCO) and different models (including Detectron-X101 and Faster-RCNN), ObjectLab consistently detects annotation errors with much better precision/recall compared to other label quality scores.