Original Source Here
- Planet led the RapidAI4EO consortium to advance land monitoring applications in Europe.
- The consortium received a Horizon 2020 grant to develop improved AI processes and training data.
- They released one of the largest training datasets of satellite imagery for machine learning research.
- The dataset is accessible on Source Cooperative, a cloud-based data publishing utility.
- It covers 500,000 patch locations across Europe, with updates every five days over two years.
- The data is sourced from Vision Impulse and Planet Fusion Monitoring, providing high-resolution and multi-sensor data.
- The dataset enables analysis of land use and cover change, benefiting research initiatives like agricultural monitoring.
- Radiant Earth praised the project for accelerating the development of Source Cooperative.
- The data has already facilitated the creation of AI-powered change detection models.
- It can unlock new machine-learning models to understand land covers and changes.
- The dataset supports Europe’s progress in tackling climate change and advancing the UN SDGs.
Main AI News:
In January 2021, Planet embarked on a mission to spearhead the RapidAI4EO consortium, propelling the advancement of cutting-edge, continuous land monitoring applications across Europe. This ambitious initiative secured a coveted grant under the prestigious Horizon 2020 program, aiming to refine AI processes and furnish essential training data for more frequent updates on land use and land cover.
Today, we take immense pride in unveiling the culmination of this program—an unparalleled training dataset of satellite imagery that is both expansive and timely, tailored to accommodate a wide array of research applications in the realm of machine learning. This exceptional dataset is readily accessible to the entire remote sensing community through Source Cooperative, Radiant Earth’s innovative cloud-based neutral data publishing utility (subject to license terms).
Encompassing a staggering 500,000 patch locations spanning Europe, with updates occurring every five days over a span of two years, this dataset ensures comprehensive country representation and spatial distribution. The Earth observation (EO) data originates from Planet’s esteemed partner, Vision Impulse, renowned for creating monthly cloud-free Sentinel-2 image mosaics boasting an impressive 10-meter resolution.
Additionally, Planet Fusion Monitoring contributes a 3-meter image every five days, amalgamating various types of multi-sensor data into a singular, uninterrupted data stream. By fusing our high-frequency, daily satellite data with publicly sourced satellite data, Fusion Monitoring unlocks a trove of insights devoid of any gaps, making it ideal for in-depth time-series analysis.
“Thanks to initiatives like Copernicus and Horizon, Europe already boasts a world-class downstream EO services industry. We firmly believe that the introduction of this groundbreaking dataset will bolster the EU’s endeavors in combating climate change, advancing the United Nations Sustainable Development Goals (UN SDGs), and fostering further growth within the European EO ecosystem,” expressed Massimiliano Vitale, Planet’s Senior Vice President of Operations EMEA.
To effectively train models in recognizing changes in landscape types, such as crops, forests, and urban regions, an abundance of time-series data proves critical. Certain land cover changes can only be discerned by observing their temporal evolution, such as seasonal crop patterns.
Although European land cover datasets have existed for some time, the introduction of this high-cadence time series, encompassing all locations, represents a pivotal innovation that equips the European region with unparalleled insights for classifying and evaluating its ever-changing landscape. While the primary focus of this dataset lies in analyzing land use and land cover changes, the invaluable insights derived from it can be generalized to numerous research initiatives that would benefit from dense time-series data, including agricultural monitoring.
“We take immense pride in serving as the host for one of the most extensive open Earth observation training datasets to date, courtesy of the RapidAI4EO consortium led by Planet,” announced Jed Sundwall, Executive Director at Radiant Earth. “The sheer magnitude of this project has accelerated the development of Source Cooperative, our state-of-the-art data publishing utility. Planet has truly established a new benchmark for open Earth observation training datasets, and we anticipate that this extraordinary dataset will enable reproducible scientific research for years to come.“
The release of this expansive training dataset of satellite imagery, coupled with advanced AI processes, signifies a significant leap forward for the market. The availability of such a comprehensive and high-frequency dataset empowers businesses and researchers in the field of land monitoring and analysis. It opens doors for the development of sophisticated machine learning models, allowing for more accurate identification and understanding of land use and cover changes.
This wealth of information enables businesses to make informed decisions, implement effective strategies, and contribute to key initiatives such as climate change mitigation and sustainable development. The market stands to benefit from the transformative insights and opportunities that arise from this remarkable collaboration between Planet, the RapidAI4EO consortium, and their partners.
Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot