Skip to content

MIT-AI-Accelerator/eie-directory

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

50 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Earth Intelligence Engine: Info Directory

Purpose: Large scale computing,  large disk storage,  GPU access.

Description: The SuperCloud allows for large batch processing that can be done in parallel, and access to dual v100 gpu nodes

Link:supercloud

Purpose: Large scale gpu computing, large ram and nvlink/infiniband support

Description: The satori cluster allows for large scale gpu batch processing.

Link: satori

Point of Contact: [email protected]

Sevir Data Generator

Over 10k spatially and temporally aligned image sequences from 5 weather sensors, including visible and infrared satellite, weather radar, and the geostationary lightning mapper.  Many samples in SEVIR correspond to NOAA’s Storm Event Database. For more info see (sevir.mit.edu).

Size: Approximately 1TB.  Total number of videos: 76k   Total number of image frames across all videos:   3.7 million

Location: Available on MIT Supercloud system (requires you to be a member of the EarthIntelligence group).  Located in EarthIntelligence/datasets/SEVIR

Link:  Not yet available.   We plan to upload to AWS in summer 2020.

Other info: Dataset tools located in sevir.

Point of Contact: [email protected]

Tropical Cyclone Idai

Images of Earth ranging from 250m resolution to 3m resolution and with frequency bands ranging from RGB-IR to 36 bands.

Size: many Terabytes.

Location: AWS, NASA, Planet, and other locations

Link:worldview

Point of Contact: [email protected]

This dataset contains high-resolution aerial imagery from the USDA NAIP program, high-resolution land cover labels from the Chesapeake Conservancy, low-resolution land cover labels from the USGS NLCD 2011 dataset, low-resolution multi-spectral imagery from Landsat 8, and high-resolution building footprint masks from Microsoft Bing, formatted to accelerate machine learning research into land cover mapping.

Size: Nearly 1 TB.

Link: landcover_data

Point of Contact: [email protected]

DeepFakes

Inputs: Image Data

Outputs: Image Data

Description: GAN for generating synthetic images from input data distribution, can be used to mix "styles", e.g. winter to summer.

Location:  several branches have been made to support running on supercloud, satori, and local clusters. Weights coming soon.

Link(s):stylegan2 , imagery_utils

Point of Contact: [email protected]

Inputs: 1-m resolution satellite imagery.

Outputs: 5-class segmentation map.

Description: U-Net architecture.

Link: landcover_model

Point of Contact: [email protected]

About

Earth Intelligence Engine Project Directory

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published