Quick answer: Run an ML backend server that implements /predict and /train, point Label Studio at its URL in project settings, and enable auto-annotation. Predictions appear as editable boxes.

The loop pays off fast. Correcting a rough box takes seconds; drawing one from scratch takes a minute. On a 5,000-image set that difference is weeks.

How does the ML backend work?

You stand up a small HTTP service with two endpoints. Label Studio calls /predict to pre-label imported tasks and /train after you finish a batch. The predictions come back in Label Studio’s own result JSON.

from label_studio_ml.model import LabelStudioMLBase

class MyModel(LabelStudioMLBase):
    def predict(self, tasks, **kwargs):
        preds = []
        for t in tasks:
            box = self.model.detect(t["data"]["image"])
            preds.append({"result": [{
                "from_name": "label", "to_name": "image",
                "type": "rectanglelabels",
                "value": {"x": box.x, "y": box.y,
                          "width": box.w, "height": box.h,
                          "rectanglelabels": [box.cls]}
            }]})
        return preds

Coordinates are percentages of image size, not pixels. Getting that wrong is the classic first bug — boxes land in the corner.

How do you connect it to a project?

1. Run the backend with label-studio-ml start ./my_backend.
2. In your project, open Settings, then Model.
3. Add a model, paste the backend URL.
4. Enable the model and turn on auto-annotation.

Verify with a validation call from the settings screen before importing thousands of tasks.

How do you close the retraining loop?

After annotators finish a batch, call /train or trigger it on a schedule. The backend pulls the latest labeled tasks, fine-tunes, and serves the new weights on the next /predict. Accuracy improves as the model sees corrected examples it got wrong before.

What infrastructure do you need?

A handful of annotators can share one small server. Inference at batch scale wants a GPU, and that is where cost appears.

What does it cost to run?

Label Studio open source is free. The ML backend is your own code, so cost equals compute — a GPU instance is typically $0.5-$2 per hour, and a model that pre-labels in under a second per image keeps that cheap. Retraining frequency is the main lever on spend.

What goes wrong?

• Coordinate systems: pixels instead of percentages puts every box in the wrong place.
• Label mismatches: the class names in your model must match the XML config exactly.
• Stale models: forgetting to trigger /train means predictions never improve.
• Confirmed bias: if annotators accept good boxes too quickly, errors in the model’s blind spots never get labeled.

Who this suits: teams with an initial model already and a growing dataset. Who should skip it: if you are starting from zero and have no model yet, label a first batch by hand; there is nothing to pre-label with.

Quick Summary

• An ML backend pre-labels tasks so humans correct instead of draw.
• Implement /predict and /train and point Label Studio at the URL.
• Use percentage coordinates, and match class names to the config exactly.
• Retraining on corrected data is what actually improves accuracy.
• Open source is free; GPU inference is the recurring cost.

Quick answer: Install with pip install label-studio, run label-studio start, create a project, pick the Image Object Detection template, and import your files. The open-source edition is free and self-hosted.

Labeling is the part of ML that no one enjoys and everyone underestimates. Getting the tool right saves hours per dataset.

How do you install and start Label Studio?

pip install label-studio
label-studio start

It opens on http://localhost:8080. The first run asks you to create an account, which stays local to your machine. There is no cloud requirement in the open-source version.

How do you configure an image labeling project?

Create a project, then pick a template. For object detection the Image Object Detection template gives you a rectangle tool and an image canvas. The labeling config is XML:

<View>
  <Image name="image" value="$image"/>
  <RectangleLabels name="label" toName="image">
    <Label value="person" background="#FFA39E"/>
    <Label value="vehicle" background="#D4380D"/>
  </RectangleLabels>
</View>

Each Label becomes a button in the UI. Keep the list short at first — ten labels is manageable, forty is where annotators start disagreeing.

How do you import data?

Upload files directly, or point at a cloud bucket with pre-signed URLs. For large image sets you want the bucket route; uploading gigabytes through the browser is slow and fragile.

[{"image": "s3://bucket/img001.jpg"}, {"image": "s3://bucket/img002.jpg"}]

Import a JSON list like this and Label Studio pulls each image into the labeling view.

How do you export annotations?

Export as JSON, COCO, YOLO, or CSV. COCO and YOLO are the two that most training frameworks read without conversion, so pick based on your model repo.

Is Label Studio free?

The open-source version is free under Apache 2.0 and runs on your own hardware. Label Studio Enterprise is a paid product with project management, quality control, and SSO. For a small team the free edition is usually plenty.

How do you keep annotation quality up?

• Write a labeling guide with example images for each class before anyone starts.
• Label 20 items yourself first; you will fix the class definitions once reality hits.
• Use overlap — assign the same image to two annotators and measure agreement.
• Review disagreements rather than majority-voting them into the training set.

What are the limits?

Self-hosting means you own the server, backups, and upgrades. The free edition lacks the built-in quality dashboards of the paid tier, so you track agreement yourself. Very large image sets need a proper database and object storage behind it, not the default SQLite and local files.

Who this suits: teams labeling their own data who want to keep it in-house. Who should skip it: if you would rather pay an outsourced labeling service and never touch the tooling, this adds work you do not want.

Quick Summary

• Label Studio is a free, self-hosted, open-source data labeling tool.
• Install with pip install label-studio and run label-studio start.
• Image Object Detection template uses an XML config with RectangleLabels.
• Export to COCO or YOLO to feed training frameworks directly.
• Budget time for a labeling guide and overlap checks — quality beats speed.