Combating Noisy Labels in Object Detection Datasets

The quality of data annotations strongly affects DNN performance, especially in complex tasks like object detection. Handling noisy annotations is often limited to accepting that some fraction of examples are incorrect, estimating their confidence, and either assigning appropriate weights or ignoring uncertain ones during training. In this work, we propose a different approach motivated by the requirements from automotive quality and safety standards. We introduce the Confident Learning for Object Detection algorithm for assessing the quality of each label in object detection datasets, identifying missing, spurious, mislabeled, and mislocated bounding boxes and suggesting corrections. By focusing on finding incorrect examples in the training datasets, we can eliminate them at the root. Suspicious bounding boxes can be reviewed to improve the quality of the dataset, leading to better models without further complicating their already complex architectures. The proposed method is able to point out nearly 80% of artificially disturbed annotation bounding boxes with a false positive rate below 10%. Cleaning the datasets by applying the most confident automatic suggestions improved mAP scores by 16% to 46%, depending on the dataset, without any modifications to the network architectures. This approach shows promising potential in rectifying state-of-the-art object detection datasets. The source code is publicly available under the GNU General Public License v3 at: https://github.com/safednn-group/safednn-clean.