README.md

metadata

datasets:
  - nkasmanoff/nasa_earth_instagram
pipeline_tag: image-to-text

Sky Scribe 🛰️✍️

This model is a fine-tuning of Microsoft's git, used to caption satellite images annotated by NASA Earth Observatory.

As a result, when prompted with satellite images that are from or resemble the images shared by NASA Earth, will return a short description of that phenomena, including guesses to the location, date, and instrument used to capture that image. These are often hallucinations, leading to an important next step in this model's development, where I can replace those guesses with additional metadata to serve as an input.

By consequence of this model being a fine-tuned version of a larger model, there is much behavior I cannot be sure is appropriate, especially for images that are not satellite images. Please use with caution.