Fig. 3
From: BAGLS, a multihospital Benchmark for Automatic Glottis Segmentation
The BAGLS dataset. (a) The folder structure of the dataset. We provide two folders for training and test, respectively, containing image/segmentation pairs, and one folder that contains the raw data (folder “raw”). (b) Exemplary images from the dataset next to the binary segmentation mask as present in the folders shown on the left. Note differing aspect ratios and image properties such as black image borders. (c) An exemplary subset from raw video 0.mp4 where two open/close cycles are visible. For illustration purposes, we cropped the image (blue frame). Consecutive frame ids (1…30) are in the lower right corner.
