Creating #geospatial understanding 🌏 from satellite images, distinguishing land use and crop types from space --> To generate an end-to-end workflow for this, I used Fused to curate the data ('fusing' USDA cropland data layer with Sentinel 2 imagery), then borrowed the Facebook Mask2Former model for fine-tuning... 🤗 the full walk-through is here on Medium: https://lnkd.in/gRWhvnSC #machinelearning #computervision
Wow, very nice write up. How long did this project take? Big fan of fused and huggingface as well.
It would be interesting to see the original satellite image in this sequence as well.
Interesting . The visualization speaks for itself 👍
whoa! game changer!
Very cool and much needed workflow streamlining. Custom pipeline development is to large scale anaylsis what wrangling is to data science: time consuming for any one individual!
This is a fantastic write up and captivating data visualizations!
Amazing read Gabe!
Thanks for sharing
Data Science and Quantum Physics
5moIt was a bugger to get the neon glow animation working in matplotlib 🤣 Full pipeline code is linked in the article https://github.com/gdurkin/geospatial_workflows. It’s all Python, part of the beauty of Fused and 🤗 working together. Low barrier to entry for data scientists. ✅