You can also store your files on your own private computer and skip uploading completely. To store a dataset on your node you have to add a new agent to the cluster.
When to use (or not) remote import¶
There are some pros and cons compared to uploading:
- Your data is private. Nobody (including Supervisely) has access to your files
- Much faster experiments. When you run Train, Inference or DTL there is no need to download & upload files from our servers - agent can start processing right away
- Harder to share. Different accounts can have access to your files only if they have the same node shared
- Your data isn't backed up. If you delete your folder, you will lose access to your datasets forever
- Annotation takes longer. Every time you open an image from your node your browser has to load it from your computer over the internet
TL;DR: Great for a single account, not so great for group work.
How to use¶
Step 1. Place files in the
Open agent directory on your node. By default agent expects files to be imported in folder
Place your files here.
Step 2. Go to import and setup settings¶
Go to Import page and select "Import from agent node" option.
Choose a name for your future project and select the node that stores your files. Don't forget to choose the correct "Import preset".
Click "Import from node" to start import.
Import from agent node is disabled
You must add at least one node on the Cluster page
Step 3. Wait till the task completes¶
You will be moved to the "Task status" page. It has a progress bar. A new project will be added after the import process is finished. No images are being uploaded, only meta information.
You can clean
Import doesn't automatically remove files from the
import folder even after the successful import. You can delete those files anytime - your files are now in