Cleanlab` is an open-source method for estimating annotation quality in semantic segmentation datasets, which are crucial in applications like robotics and medical imaging. It automatically detects mislabeled images by scoring each image based on its overall labeling quality and estimating which images contain annotation errors. This allows users to prioritize which images should be reviewed first, avoiding training models with bad data. The method has been benchmarked against other error detection algorithms and shown to detect mislabeled images with higher precision and recall. It can be used with any segmentation model trained via any strategy, making it a valuable tool for improving the quality of datasets in AI applications.