Pretty quiet in here
Machine learning for audio engineering
| Website | http://www.christiansteinmetz.com |
| GitHub | https://github.com/csteinmetz1 |
| Website | http://www.christiansteinmetz.com |
| GitHub | https://github.com/csteinmetz1 |
Let's continue the #discussion but with a few new clarifications...
Can a self-supervised neural representation that maps audio/spectrograms to fixed dim embeddings at a fixed temporal resolution *simultaneously* solve a range of existing audio tasks (e.g the HEAR Benchmark)?
If so, why hasn't this worked yet? If not, what else do we need?