Hey @for_east,
Welcome to the community. There have been a great many discussions in the Deep Learning Specialization in the past, regarding what to do in the absence of any human-level performance. Here are some posts for your reference:
- Bayes error, human-level performance and overfitting (structured data)
- Human Level Performance, how to set it?
- What to do when there is no human-level performance baseline?
- How to measure human level performance against human made labels?
You will find that these posts share a great deal of knowledge regarding your query, so let me know, if these help.
Cheers,
Elemento