fiftyone-dataset-inference

Run model inference by setting your dataset context, launching the FiftyOne App, and executing the apply_zoo_model operator with a chosen zoo model and label field to populate predictions.

FiftyOne zoo models support inference for computer vision tasks including object detection, classification, segmentation, and embeddings, allowing you to apply these models directly to your dataset and output labeled prediction fields.

Yes, applying zoo models requires a running FiftyOne App and an existing dataset. You must set the context and launch the app before executing the inference operator to ensure predictions are generated correctly.

Yes, you can build end-to-end workflows by importing data first and then running model inference in a single flow. This allows you to seamlessly apply zoo models to newly imported datasets and view the results immediately.

To view model predictions alongside ground truth, run inference to create a labeled prediction field, then use set_view with an exists filter to inspect the generated predictions and compare them directly with existing labels.