Machine Learning and CNNs on Imagery

Meta Trend	Big Data
Title	Machine Learning and CNNs on Imagery
Description	Machine learning is the subfield of computer science that gives computers the ability to learn without being explicitly programmed. Deep learning and Convolutional Neural Networks (CNNs) - a sub type of machine learning - consists of multiple hidden layers in an artificial neural network - (Wikipedia).
What is new or emerging?	Deep Learning is advancing rapidly based on increasing computing capabilities, image databases and trained CNNs.
Why might it matter?	The ability to detect features using CNNs
Horizon	Next: Multiple CNNs have been applied to remote sensed data with impressive results. Questions remain about provenance and uncertainty. There is a need for trained models. Transferability of trained models and interoperability of CNNs are open questions.
Impact	Disruptive:
Gartner Hype Curve phase
Technology Readiness Level
Discussion Issue	Discussion of Trend on GitHub
References	1. ImageNet Classification with Deep Convolutional Neural Networks 2. Big Geospatial Data – an OGC White Paper 3. NYC task force on automated decision systems used by agencies 4. Team Breaks Exaop Barrier With Deep Learning Application
Examples	URL to technology implementation examples
Geospatial Tech Category
OGC Working Groups	Big Data DWG, OGC Testbed 14 includes work on Machine learning.

Provide feedback