Language: 日本語English

fiftyone Plugin

Topics
AI Agents & AI App Development · Data Engineering & Analytics · Design & Media Production
First seen
2026-07-09
Last confirmed
2026-07-13
Explanation last updated
2026-07-12

The explanation below is AI-generated. Please verify it against the sources.

FiftyOne is a tool for building high-quality datasets and computer vision models — it lets you visualize datasets, analyze models, find duplicates, run inference, evaluate predictions, and develop custom plugins, according to the FiftyOne documentation. The linked source repository, fiftyone-skills, packages this functionality into "Skills": step-by-step workflows that teach AI assistants (such as Claude Code, Cursor, or Gemini CLI) to perform FiftyOne tasks autonomously via natural language, per the fiftyone-skills README. Skills bridge natural language and FiftyOne's 80+ operators, and MCP-dependent skills require the separate FiftyOne MCP Server to interact with datasets and run operators. The repository lists 18 skills covering dataset import/export, duplicate detection, model inference, evaluation, embeddings visualization, plugin development, and more. Installation is done via a universal installer script or agent-specific commands (e.g., Claude Code plugin install, Gemini CLI extension install).

Overview

FiftyOne is a tool (per docs.voxel51.com) for building and curating computer vision datasets, visualizing data and model predictions in an interactive app, evaluating models, and developing plugins. The fiftyone-skills repository packages this functionality into "Skills" that AI coding assistants load to carry out FiftyOne workflows on request, working alongside the separate FiftyOne MCP Server for skills that need to interact with live datasets.

What you can do with fiftyone

  • Install Skills via a universal installer script or agent-specific commands (Claude Code, Gemini CLI, and others)
  • Import and export datasets in formats such as COCO, YOLO, VOC, CVAT, CSV, and Hugging Face Hub
  • Find and remove duplicate images using brain similarity
  • Run Zoo models for detection, classification, segmentation, and embeddings
  • Compute model evaluation metrics such as mAP, precision, recall, and confusion matrices
  • Visualize embeddings in 2D to find clusters and identify outliers
  • Develop custom FiftyOne plugins (operators and panels)
  • Curate datasets end-to-end, including quality checks, annotation audits, class distribution, and splits
  • Troubleshoot common issues such as dataset persistence, App connection problems, and MongoDB errors

Sources

Original description (English)

Build high-quality datasets and computer vision models. Visualize datasets, analyze models, find duplicates, run inference, evaluate predictions, and develop custom plugins.

History of fiftyone

Back to list