deep learning (TensorFlow) + Google Earth Engine

[bbest]

This is the future!

• SpaceKnow.com

  

  • Google Cloud Platform Blog: Powering geospatial analysis: public geo datasets now on Google Cloud

  • Planet Labs satellite imagery data is now on Google Cloud

• Terrapattern can identify things like:

  

• jldowns/google_earth_engine_notebook: a walkthrough of some Google Earth Engine Features, as well as using the data in TensorFlow with Docker instantiation

• Frontiers | Exploring Google Earth Engine Platform for Big Data Processing: Classification of Multi-Temporal Satellite Imagery for Crop Mapping | Earth Science

• Google Earth Engine: Planetary-scale geospatial analysis for everyone - ScienceDirect

• Planet: Understanding the Amazon from Space | Kaggle

• See my Deep Learning Group Notes - Google Docs

[bbest]

tensorflow + shiny

Mini AI app using TensorFlow and Shiny – Opiate for the masses

[bbest]

developmentseed: segnet + satellite imagery

Presentation:

• GISDay@Stanford 2017: End-to-End Machine Learning with Satellite Imagery

The open source libraries for machine learning have dramatically improved over the past two years. However the related tools for managing a machine learning process are still lacking. In particular, there are few good tools for collecting and preparing training data; vectorizing and stitching computer vision outputs; and cleaning and reprocessing.The Skynet suite is an end-to-end set of tools for extracting useful data from a machine learning process. Skynet is optimized for feature detection from satellite and drone imagery. At its core Skynet is an application of Segnet, a convolutional neural network approach to semantic segmentation. We've built a suite of tools around Skynet to collect and manage training data, inventory trained models, produce useful vectorized data outputs, and to optimize cleaning of that data. The Skynet suite primarily leverages the OSM ecosystem for training data. Skynet-data ingests data from OSM and prepares it for use as training data. Skynet-collect is a set of OSM based data collection tools. Skynet-scrub is used by data scrubbers to quickly clean and improve the outputs of the computer vision process. Data created by Skynet can be pushed directly into an private-OSM instance, staged for inclusion in global-OSM, or exported for use in another tool or app.

Articles:

• End-to-End Satellite Imagery Analysis — Development Seed
• Machine Learning at State of the Map US — Development Seed
• SegNet

Github repos at github.com/developmentseed:

• developmentseed/skynet-data: Data pipeline for machine learning with OpenStreetMap
• developmentseed/skynet-train: Training and test the SegNet neural network on satellite imagery
• developmentseed/skynet-scrub: GUI for editing machine learning outputs
• developmentseed/skynet-scrub-server: Backing store for developmentseed/skynet-scrub
• developmentseed/landsat-util: A utility to search, download and process Landsat 8 satellite imagery
• developmentseed/climatescope-data: The data processing scripts for the Global Climatescope