Skip to content

Latest commit

 

History

History
34 lines (26 loc) · 1.21 KB

README.md

File metadata and controls

34 lines (26 loc) · 1.21 KB

Rareplane datasets

Datasets

Rareplanes consists in two datasets: one dataset of observed satellite images ("real") and one dataset of computer generated images ("synthetic").

S3 structure

The structure of the s3 bucket rareplanes-public is detailed in the user guide available there.

Prerequisites

  • install the aws cli
  • run aws configure

Download datasets

Download all the data using the command below. Both real and synthetic datasets account for over 500 GB of data, so this command will take a while.

python download_data.py --dataset_type real synthetic --split train test

If you only need data for training object detection algorithms, the command below will only pull the 512x512 observed tiles and the synthetic images.

python download_data.py --dataset_type real synthetic --split train test --training_only

You can also download any data of interest by using directly the AWS CLI. For example, the command below will only download the full resolution satellite images:

aws s3 sync s3://rareplanes-public/real/train/PS-RGB/ real/train/PS-RGB/