search

joonaskalda / PixIT

Companion repo for the paper "PixIT: Joint Training of Speaker Diarization and Speech Separation from Real-world Multi-speaker Recordings" published at Odyssey 2024
27 stars 1 forks source link

Will there be general domain / Domain agnostic models? #1

Open asusdisciple opened 3 months ago

asusdisciple commented 3 months ago

Since the current model is solely trained on the AMI corpus, do you have any intentions to evaluate and train in a broader context? Having a general model in that sense would boost its popularity immensly I think and open it up for a lot of more use cases.

joonaskalda commented 3 months ago

Thank you for your interest @asusdisciple! Training on a larger dataset and improving robustness are indeed things we plan to experiment with.