Benchmark for Automatic Glottis Segmentation (BAGLS)
Welcome to the website of the Benchmark for Automatic Glottis Segmentation (BAGLS). BAGLS is the first large-scale, publicly available dataset of endoscopic high-speed video with frame-wise segmentation annotations. It has been collected in a collaboration of seven institutions and features a total of 59,250 frames. Usage notes of the dataset are found in the BAGLS publication (Gómez, Kist et al., Sci Data 2020). To download the complete dataset please use the Zenodo or Kaggle website. Below you can inspect individual frames with the corresponding segmentation masks and individual metadata for the frames, as well as the videos provided with BAGLS. The raw data is on the Zenodo archive.