Dataset Inference#
Run ML model inference on a FiftyOne dataset using Zoo models or your own. Supports detection, classification, segmentation, and embeddings.
Install#
curl -sL skil.sh | sh -s -- voxel51/fiftyone-skills
When prompted, select fiftyone-dataset-inference from the menu.
Requirements#
FiftyOne MCP Server (optional, recommended for operator-based inference)
Usage#
Load a dataset in FiftyOne, then ask your AI assistant:
"Run YOLO object detection on my dataset"
"Compute CLIP embeddings for all images"
"Apply a segmentation model and store results in a new field"
The skill discovers available models dynamically, confirms the model and output field with you, then runs inference.
Example#
import fiftyone as fo
import fiftyone.zoo as foz
dataset = foz.load_zoo_dataset("quickstart")
Then ask your assistant:
"Run object detection on the quickstart dataset using a YOLO model"