Open-Source Manually Annotated Vocal Tract Database for Automatic Segmentation from 3D MRI Using Deep Learning: Benchmarking 2D and 3D Convolutional and Transformer Networks

This work addresses the need for efficient and accurate vocal tract segmentation for voice and speech applications, but appears incremental as it benchmarks existing methods on a new dataset.

The study tackled the problem of time-intensive and error-prone manual segmentation of the vocal tract from 3D MRI data by evaluating deep learning algorithms for automatic segmentation, benchmarking 2D and 3D convolutional and transformer networks.