...the who's who,
and the what's what 
of the space industry

Space Careers

news Space News

Search News Archive

Title

Article text

Keyword

  • Home
  • News
  • Satellogic unveils expansive high-resolution image dataset for AI training

Satellogic unveils expansive high-resolution image dataset for AI training

Written by  Friday, 03 May 2024 19:59
Write a comment
Los Angeles CA (SPX) May 03, 2024
Satellogic Inc. (NASDAQ: SATL) has made public a substantial open dataset from its collection. This dataset is composed of approximately 3 million unique location images, doubling to 6 million with revisits, and is designed to enhance the training of AI foundation models. Each 384x384 pixel image contributes to the overall 900 Gigapixels, covering diverse land uses, objects, geographies, and sea
Satellogic unveils expansive high-resolution image dataset for AI training
by Clarence Oxford
Los Angeles CA (SPX) May 03, 2024

Satellogic Inc. (NASDAQ: SATL) has made public a substantial open dataset from its collection. This dataset is composed of approximately 3 million unique location images, doubling to 6 million with revisits, and is designed to enhance the training of AI foundation models. Each 384x384 pixel image contributes to the overall 900 Gigapixels, covering diverse land uses, objects, geographies, and seasonal variations. This freely accessible dataset is available on Hugging Face.

"Following a stream of recent publications, with the release of this large dataset we aim to accelerate the development of foundational models in the field of EO," said Javier Marin, Applied AI Director at Satellogic. "Instead of relying on analysts to manually select and process satellite images, we will soon start interacting with large Earth Observation AI models with access to high-resolution, real-time imagery of our planet to derive those insights."

The data, released under the Creative Commons CC-BY 4.0 license, permits commercial utilization with proper attribution.

Accompanying the dataset's release, Satellogic will also introduce a baseline foundation model, a masked autoencoder, notable for its scalable self-supervised learning capabilities in computer vision. The forthcoming paper details the dataset's composition, model framework, and experimental approaches, showcasing the collaborative efforts of a prominent research team including Alexandre Lacoste at ServiceNow, guided by Yoshua Bengio

Related Links
Satellogic
Earth Observation News - Suppiliers, Technology and Application


Read more from original source...

You must login to post a comment.
Loading comment... The comment will be refreshed after 00:00.

Be the first to comment.

Interested in Space?

Hit the buttons below to follow us...