SEAM AI team held the "Seismic Facies Identification Challenge" a while ago, where all the labeled data is provided as a part of this challenge. The New Zealand Parihaka dataset was manually interpreted to represent six unique facies. Being involved in a similar development, I can not stress the difficulty of manually picking all the facies. Kudos to Chevron!
Even though the challenge is over, you can register and access the labeled dataset and the participants' solutions.
If you would like to give it a try to beat the highest score, or learn some basics of the problem and the tooling, visit the "Notebook" section. There are several good explainers notebooks with code, and some of them on Google Colab.
Consider checking out my other post about facies classification:
◼️ Seismic facies prediction code by Microsoft -https://lnkd.in/dpf9_qWK
◼️ Make AI follow geological knowledge and interpretation rules (part 1 - facies analysis) - https://lnkd.in/dJ87DtA
◼️ Seismic Facies identification with RNN claims better results than CNN - https://lnkd.in/dmji7Kn7