Moonshine#
Moonshine is a Python library that makes it easy for remote sensing researchers,professionals, and enthusiasts to develop ML models on their data. It providespre-trained models across a variety ofdatasetsand architectures, allowing you to reduce your labeling costs and compute requirementsfor your own application.
Why Use Moonshine?#
Pretrained on multispectral data: Many existing packages are pretrained withImageNet or similar RGB images. Using Moonshine you can unlock the full power ofsatellites that many contain many channels of multispectral data.
Pretrained on remote sensing data: Pretraining in the domain of your data isimportant, and most off the shelf pretrained models are fit to natural images such asImageNet.
Focus on usability: While there are some academic remote sensing pretrainedmodels available, they often are difficult to use and lack support. Moonshine isdesigned to be easy to use and will offer community support via Github and Slack.
Need more convincing that Moonshine works? Check out this comparison of Moonshinepretrained weights vs training from scratch:
The above chart shows the difference between training thefunctional map of the world classification task usingour pre-trained model vs. training from scratch. The task is to classify patches ofsatellite data by the functional purpose of the land, with 63 possible classes and over300,000 training images.
Training from scratch both performs worse overall, and for roughly the same level ofaccuracy we can train for 45% less time (28h vs 16h on a V100). Check out thequick start section for further information, includinghow to install the library.