Noisy Annotation Refinement for Object Detection
This work addresses the costly and noisy annotation issue in object detection, particularly for datasets from economical sources like crowdsourcing, offering a solution to improve training efficiency and accuracy.
The paper tackles the problem of training object detectors on datasets with noisy annotations, including inaccurate bounding boxes and incorrect class labels, by proposing a method that decouples and corrects these noises, resulting in significant performance improvements over baselines across various noise levels.
Supervised training of object detectors requires well-annotated large-scale datasets, whose production is costly. Therefore, some efforts have been made to obtain annotations in economical ways, such as cloud sourcing. However, datasets obtained by these methods tend to contain noisy annotations such as inaccurate bounding boxes and incorrect class labels. In this study, we propose a new problem setting of training object detectors on datasets with entangled noises of annotations of class labels and bounding boxes. Our proposed method efficiently decouples the entangled noises, corrects the noisy annotations, and subsequently trains the detector using the corrected annotations. We verified the effectiveness of our proposed method and compared it with the baseline on noisy datasets with different noise levels. The experimental results show that our proposed method significantly outperforms the baseline.