On the Supervise.ly platform, navigate to the concerned Workspace and select the project to work on. Now, download the data as a zipped archive, choosing to download the images along with the annotations.
Supervise.ly will then run a batch job, generating the necessary .zip archive. Once it’s downloaded, extract using a standard archiver. To do so using the command line, open a terminal prompt and navigate to the download path. Now, type in the following command:
unzip -q DOWNLOADED_FILE.zip -d SAVE_PATH
where DOWNLOADED_FILE.zip is the name of the downloaded file and SAVE_PATH is the destination on the disk.
You’re now ready to generate the COCO-style dataset records expected by the neural network.