Skip to main content

How-To Guides

How to Create a Segmentation Pipeline​

  1. From the Dashboard, click + New Pipeline.
  2. Enter a Pipeline Name and a valid SLURM Account.
  3. Optionally add a Description.
  4. Click Create & Open — the Image Annotator canvas opens immediately.

How to Upload Images to the Cluster​

  1. From the Dashboard, click Upload Data.
  2. Select a Target System from the dropdown (e.g. Pitzer (OSC)).
  3. Enter or browse to an absolute Destination Path on that system.
  4. Drag and drop files into the upload zone, or click Select Files / Select Directory.
  5. Click Upload to transfer.

How to Use Smart Click (SAM3-Assisted Masking)​

  1. Open an image in the annotation canvas.
  2. Click the Smart Click tool (magic wand / wand-with-sparkle icon) in the toolbar.
  3. Select Single Click mode and enter a label, then click on an object to generate its mask automatically.
  4. Optionally enable SAHI and adjust Crop Size for better coverage of small objects in large images.
  5. Click Enter to apply.

How to Use Text Prompt Annotation​

  1. Open an image in the annotation canvas.
  2. Click the Smart Click tool and switch mode to Text Prompt.
  3. Enter a comma-separated list of class names (e.g. zebra, giraffe).
  4. Optionally enable SAHI for improved small-object recall.
  5. Click Enter — SAM3 generates masks for all detected instances of those classes.

How to Flag Masks for Review​

  1. In the right panel, locate the mask in the Masks list.
  2. Click the flag icon next to a mask entry to mark it as Needs Review.
  3. Use the Filter by Flag section at the top of the right panel to show only flagged masks.
  4. Add custom flag names using the New flag name... field and the + button.

How to Filter Masks by Label or Flag​

  • Click any label chip in the Filter by Label section to show only masks of that class.
  • Click any flag chip in the Filter by Flag section to show only masks with that flag.
  • Clicking an active chip again deselects it and shows all masks.

How to Remove Low-Confidence Masks​

  1. Adjust the Confidence slider in the right panel to the desired threshold.
  2. Click Remove masks below score: X.XX to permanently delete all masks under that threshold.

How to Save Annotations to the Cluster​

  1. Click the Save (floppy disk) icon in the toolbar.
  2. Select a cluster system from the dropdown.
  3. Enter the full remote File Path (e.g. /path/to/save/annotations.json).
  4. Toggle COCO JSON on or off to select the output format.
  5. Click Save.

How to Import Existing Annotations​

  1. Click the Import (upload arrow) icon in the toolbar.
  2. Select the cluster from the dropdown or click Browse File to upload from your local machine.
  3. Enter the remote path in Or Enter File Path if the file is already on the cluster.
  4. Toggle the format switch to match your file (COCO JSON or Default JSON).
  5. Click Upload.

How to Download Annotations Locally​

Click the Download icon in the toolbar and select COCO JSON or Default JSON. The file is saved directly to your local machine.

How to Resume a Segmentation Pipeline​

From the Dashboard, find the pipeline by name or ID using the search bar. Click Open Pipeline on the pipeline card to return to the Image Annotator at the current state.

How to Set Up Hugging Face Access​

Some models used by Smart Labeler (SAM3) are gated on Hugging Face and require an account, approved access, and a personal token. Complete all four steps in order.

Step 1 — Create a Hugging Face Account​

Go to huggingface.co and click Sign Up. Enter your email, choose a username and password, and verify your email address. Complete your profile (name, organization if applicable).

Tip: Use a professional or institutional email address — it improves your chances of approval for gated models. If you already have an account, skip to Step 2.

Step 2 — Request Access to SAM3​

SAM3 is a gated model hosted by Meta. You must agree to their usage terms before you can use it.

  1. Open the SAM3 model page: huggingface.co/facebook/sam3
  2. Click Request Access and fill in your details (first name, last name, date of birth, country, affiliation, job title).
  3. Accept the Meta Privacy Policy terms and click Submit.
  4. You will see a confirmation that your request is pending review.
  5. Once approved, the model page shows: "Gated model — You have been granted access to this model."

Tip: You will receive an email when approved. Check your spam folder if you don't see it after 24 hours.

Step 3 — Generate a Hugging Face Access Token​

Once your access request is approved, generate an API token so Smart Labeler can authenticate model downloads on your behalf.

  1. Navigate to your token settings: huggingface.co/settings/tokens
  2. Click + Create new token.
  3. Set the Token Type to Read and give it a descriptive name (e.g., ICICLE-TapisAccess).
  4. Click Create token and copy the value immediately — it is only shown once.

Warning: Treat your token like a password. Do not share it or commit it to a repository. If you lose it, revoke it and generate a new one from the same settings page.

Step 4 — Add the Token in ICICLE / Tapis​

With your token ready, add it to your ICICLE/Tapis credentials so it can be injected securely into HPC jobs.

  1. Log in to your ICICLE/Tapis account.
  2. Navigate to the Settings section indicated by 3 dots on the dashboard.
  3. Click on Access Key.
  4. Paste your token and save.
  5. Click on Revoke token to permanently delete the token if needed.