Annotations and download

After rendering, click on the View gallery button to access the gallery. Here, you can set up your classes and annotations and and prepare your dataset for download.

Each scenario.render() call or sent render job creates a collection of images. You can download multiple collections together as a single dataset by ticking them, then setting up the class ontology if necessary.

Export the right annotations

Click on Download to proceed and you’ll see a confirmation window.

If you haven’t set up a class ontology, click on Edit class ontology to ensure that the right annotations are exported for your dataset.

Assign assets to classes by dragging them from the left panel (will not be labeled) to the right panel (will be labeled). You can create new classes by clicking on the + Add new class button.

Note

If you don’t set up a class ontology, the dataset will be exported without annotations.

You may leave assets unlabeled, which is useful for teaching vision models to ignore background objects or distractors.

Downloading the dataset

After setting up the class ontology, click on the Download button to download the dataset. Depending on the size of the dataset, it may take a few minutes for the dataset and annotations to be prepared. Once ready, downloads will begin automatically as a ZIP file!