FiftyOne provides a powerful plugin framework that allows for extending and customizing the functionality of the tool.
With plugins, you can add new functionality to the FiftyOne App, create integrations with other tools and APIs, render custom panels, and add custom buttons to menus.
With FiftyOne Teams, you can even write plugins that allow users to execute long-running tasks from within the App that run on a connected compute cluster.
For example, here's a taste of what you can do with the @voxel51/brain plugin!
brain.mp4
This repository contains a curated collection of FiftyOne Plugins, organized into the following categories:
- Core Plugins: core functionality that all FiftyOne users will likely want to install. These plugins are maintained by the FiftyOne team
- Voxel51 Plugins: non-core plugins that are officially maintained by the FiftyOne team
- Example Plugins: these plugins exist to inspire and educate you to create your own plugins! Each emphasizes a different aspect of the plugin system
- Community Plugins: third-party plugins that are contributed and maintained by the community. These plugins are not officially supported by the FiftyOne team, but they're likely awesome!
🔌🤝 Contribute Your Own Plugin 🚀🚀
Want to showcase your own plugin here? See the contributing section for instructions!
| Name | Tags | Description |
|---|---|---|
| @voxel51/annotation | annotation | ✏️ Utilities for integrating FiftyOne with annotation tools |
| @voxel51/brain | curation visualization | 🧠 Utilities for working with the FiftyOne Brain |
| @voxel51/dashboard | visualization | 📊 Create your own custom dashboards from within the App |
| @voxel51/evaluation | evaluation | ✅ Utilities for evaluating models with FiftyOne |
| @voxel51/io | io | 📁 A collection of import/export utilities |
| @voxel51/indexes | utils | 📈 Utilities working with FiftyOne database indexes |
| @voxel51/plugins | utils | 🧩 Utilities for managing and building FiftyOne plugins |
| @voxel51/delegated | utils | 📡 Utilities for managing your delegated operations |
| @voxel51/runs | utils | 🏃 Utilities for managing your custom runs |
| @voxel51/utils | utils | ⚒️ Call your favorite SDK utilities from the App |
| @voxel51/zoo | model dataset | 🌎 Download datasets and run inference with models from the FiftyOne Zoo, all without leaving the App |
| Name | Tags | Description |
|---|---|---|
| @voxel51/voxelgpt | examples | 🤖 An AI assistant that can query visual datasets, search the FiftyOne docs, and answer general computer vision questions |
| @voxel51/mlflow | training | 📋 Track model training experiments on your FiftyOne datasets with MLflow! |
| @voxel51/huggingface_hub | dataset huggingface | 🤗 Push FiftyOne datasets to the Hugging Face Hub, and load datasets from the Hub into FiftyOne! |
| @voxel51/transformers | model huggingface | 🤗 Run inference on your datasets using Hugging Face Transformers models! |
| @voxel51/rerun-plugin | visualization | 🎥 Visualize Rerun data files (.rrd) inside the FiftyOne App |
| Name | Tags | Description |
|---|---|---|
| @voxel51/hello-world | examples | 👋 An example plugin that contains both Python and JavaScript components |
| @voxel51/operator-examples | examples | ⚙️ A collection of example operators showing how to use the operator type system to build custom FiftyOne operations |
| @voxel51/panel-examples | examples | 📊 A collection of example panels demonstrating common patterns for building Python panels |
🔌🤝 Contribute Your Own Plugin 🚀🚀
Want to showcase your own plugin here? See the contributing section for instructions!
| Name | Tags | Description |
|---|---|---|
| @landingai/ade | ocr model text | 📄 Parse, extract, and split documents using LandingAI's Agentic Document Extraction (ADE) API. Converts PDFs, images, spreadsheets, and Office files into structured Markdown with spatial bounding box grounding. |
| @parva101/video2dataset | utils | 🎬 Convert YouTube URLs or local videos into FiftyOne image datasets with uniform/scene-change/hybrid frame sampling, perceptual deduplication, and source metadata. |
| @harpreetsahota/molmo_point | model tracking | 🫵🏻 Integrating MolmoPoint a model that locates and tracks objects in images and videos by pointing and returning precise pixel coordinates |
| @harpreetsahota/online_video_depth_anything | model depth | 🏙️ Integrating Online Video Depth Anything (oVDA) for a temporally-consistent monocular depth estimator for videos that runs in an online setting with low VRAM consumption. |
| @harpreetsahota/vlm_prompt_lab | model vlm | 👨🏽🔬 Experiment with any VLM that can be run in a Hugging Face image-text-to-text pipeline right in the FiftyOne App! |
| @ehofesmann/envi-spetral-viewer | visualization hyperspectral | 🌈 Explore hyperspectral image datasets, interactively visualize pixel-level spectra, and dynamically recolor images. |
| @harpreetsahota/qwen3_5_vl | model | 👀 Implementing Qwen3.5VL as a Remote Source Zoo Model for FiftyOne. |
| @harpreetsahota/FiftyComfy | visualization | |
| @harpreetsahota/hf_fine_tuner_plugin | model huggingface | 🎚️ A plugin to fine-tune Hugging Face models on your FiftyOne Dataset. |
| @harpreetsahota/image_editing_panel | model huggingface | 🎑 Chat-based image editing powered by HuggingFace image-to-image Inference API. |
| @harpreetsahota/qwen_image_edit | model | 🏞️ Chat-based image editing powered by drbaph/Qwen-Image-Edit-2511-FP8 |
| @mgustineli/roi-patches | curation | 🪟 Tile images into a configurable grid of ROI patches with adjustable overlap for region-based analysis, using FiftyOne's native patches view. |
| @harpreetsahota/LightOnOCR-2 | model vlm | 📑 LightOnOCR-2-1B is a compact multilingual VLM that converts document images into clean, naturally ordered text without brittle multi-stage OCR pipelines. |
| @harpreetsahota/glm_ocr | model vlm | 📄 GLM-OCR is a lightweight 0.9B vision language model achieving state-of-the-art document understanding, including formula recognition, table recognition, and structured information extraction. |
| @harpreetsahota/cradiov4 | model embeddings | 📻 CRADIOv4 performs visual feature extraction whose image embeddings can be used by a downstream model for various tasks. This implementation also produces attention maps. |
| @perceptron-ai-inc/isaac-0_2 | model vlm | 🤖 Isaac-0.2 is Perceptron AI's hybrid-reasoning vision language model supporting object detection, keypoint detection, OCR, instance segmentation, visual question answering, and UI understanding. Includes thinking and tool use for improving detection in complex scenes. |
| @harpreetsahota/medgemma_1_5 | model medical | 🩻 Implementing MedGemma 1.5 as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/qwen3vl_embeddings | model embeddings | 📼 Qwen3-VL-Embedding maps text, images, and video into a unified representation space, enabling powerful cross-modal retrieval and understanding. |
| @ardamamur/egoexor | dataset medical | 🏥 EgoExOR is an Operating Room dataset fusing egocentric and exocentric perspectives for surgical procedures. See here to load it with FiftyOne. |
| @harpreetsahota/molmo2 | model vlm | 📹 Molmo2 is a family of open vision language models developed by the Allen Institute for AI (Ai2) that support image, video, and multi-image understanding and grounding. |
| @harpreetsahota/apple_sharp | model 3d | 🧊 SHARP is Apple's state-of-the-art model for predicting 3D Gaussian Splats from a single RGB image. This integration brings SHARP to FiftyOne, enabling batch inference on image datasets with 3D visualization. |
| @harpreetsahota/sam3_images | model segmentation | 🖼️ Integration of Meta's SAM3 (Segment Anything Model 3) into FiftyOne, with full support of text prompts, keypoint prompts, bounding box prompts, auto segmentation, and image embeddings. |
| @harpreetsahota/qwen3vl_video | model vlm | 🎥 A FiftyOne zoo model integration for Qwen3-VL that enables comprehensive video understanding with multiple label types in a single forward pass and for computing video embeddings. |
| @harpreetsahota/text_evaluation_metrics | model evaluation text | 🔡 This plugin provides five text evaluation metrics for comparing predictions against ground truth: ANLS, Exact Match, Normalized Similarity, Character Error Rate, and Word Error Rate. |
| @harpreetsahota/mineru_2_5 | model vlm | 📜 MinerU2.5 is a 1.2B-parameter vision language model for efficient high-resolution document parsing. This model can support grounding OCR as well as free text OCR. |
| @harpreetsahota/nomic-embed-multimodal | model embeddings | 📜 Nomic Embed Multimodal is a family of vision language models built on Qwen2.5-VL that generates high-dimensional embeddings for both images and text in a shared vector space. |
| @harpreetsahota/bimodernvbert | model embeddings | 🗂️ BiModernVBert is a vision language model built on the ModernVBert architecture that generates embeddings for both images and text in a shared 768-dimensional vector space. |
| @harpreetsahota/colmodernvbert | model embeddings | 📑 ColModernVBert is a multi-vector vision language model built on the ModernVBert architecture that generates ColBERT-style embeddings for both images and text. |
| @harpreetsahota/deepseek_ocr | model vlm | 🐳 DeepSeek-OCR is a vision language model designed for optical character recognition with a focus on "contextual optical compression." |
| @harpreetsahota/olmOCR-2 | model ocr | 📊 olmOCR-2 is a state-of-the-art OCR model built on Qwen2.5-VL architecture that extracts text from document images with high accuracy. |
| @harpreetsahota/jina_embeddings_v4 | model embeddings | 📑 Jina Embeddings v4 is a state-of-the-art vision language model that generates embeddings for both images and text in a shared vector space. |
| @harpreetsahota/colqwen2_5_v0_2 | model embeddings | 🗃️ ColQwen2.5 is a vision language model based on Qwen2.5-VL-3B-Instruct that generates ColBERT-style multi-vector representations for efficient document retrieval. This version takes dynamic image resolutions (up to 768 image patches) and doesn't resize them, preserving aspect ratios for better accuracy. |
| @harpreetsahota/nanonets_ocr2 | model ocr | 📄 Nanonets-OCR2 transforms documents into structured markdown with intelligent content recognition and semantic tagging, making it ideal for downstream processing by Large Language Models (LLMs). |
| @harpreetsahota/colpali_v1_3 | model embeddings | 📃 ColPali is a vision language model based on PaliGemma-3B that generates ColBERT-style multi-vector representations for efficient document retrieval. |
| @harpreetsahota/kosmos2_5 | model ocr | 📑 Kosmos-2.5 excels at two core tasks: generating spatially-aware text blocks (OCR) and producing structured markdown output from images. |
| @harpreetsahota/moondream3 | model vlm | 🌝 Moondream 3 (Preview) is an vision language model with a mixture-of-experts architecture (9B total parameters, 2B active). This model makes no compromises, delivering state-of-the-art visual reasoning while still retaining our efficient and deployment-friendly ethos. |
| @harpreetsahota/caption_viewer | visualization vlm | 🖥️ A plugin that intelligently displays and formats vision language model outputs and text fields. Perfect for viewing OCR results, receipt analysis, document processing, and any text-heavy computer vision workflows. |
| @harpreetsahota/fiftyone_wandb_plugin | utils evaluation | 📉 This plugin connects FiftyOne datasets with Weights & Biases to enable reproducible, data-centric ML workflows. |
| @harpreetsahota/isaac0_1 | model vlm | 🤖 Isaac-0.1 is the first in Perceptron AI's family of models built to be the intelligence layer for the physical world. This integration supports various computer vision tasks including object detection, classification, OCR, visual question answering, and more. |
| @vlm-run/vlmrun-voxel51-plugin | model vlm | 🎯 Extract structured data from visual and audio sources including documents, images, and videos |
| @harpreetsahota/minicpm-v | model vlm | 👁️ Integrating MiniCPM-V 4.5 as a Remote Source Zoo Model in FiftyOne |
| @harpreetsahota/fast_vlm | model vlm | 💨 Integrating FastVLM as a Remote Source Zoo Model for FiftyOne |
| @harpreetsahota/gui_actor | model vlm | 🖥️ Implementing Microsoft's GUI Actor as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/synthetic_gui_samples_plugins | model vlm | 🧪 A FiftyOne plugin for generating synthetic samples for datasets in COCO4GUI format |
| @harpreetsahota/coco4gui_fiftyone | io | 💽 Implementing the COCO4GUI dataset type in FiftyOne with importers and exports |
| @harpreetsahota/fiftyone_lerobot_importer | io | 🤖 Import your LeRobot format dataset into FiftyOne format |
| @harpreetsahota/medsiglip | model medical | 🩻 Implementing MedSigLIP as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/florence2 | model vlm | 🏛️ Implementing Florence2 as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/medgemma | model medical | 🩻 Implementing MedGemma as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/moondream2 | model vlm | 🌔 Moondream2 implementation as a remotely sourced zoo model for FiftyOne |
| @harpreetsahota/qwen2_5_vl | model vlm | 👀 Implementing Qwen2.5-VL as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/paligemma2 | model vlm | 💎 Implementing PaliGemma-2-Mix as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/siglip2 | model vlm | 🔎 A FiftyOne Remotely Sourced Zoo Model integration for Google's SigLIP2 model enabling natural language search across images in your FiftyOne Dataset |
| @harpreetsahota/os_atlas | model vlm | 🖥️ Integrating OS-Atlas Base into FiftyOne as a Remote Source Zoo Model |
| @harpreetsahota/Nemotron_Nano_VL | model vlm | 👁️ Implementing Llama-3.1-Nemotron-Nano-VL-8B-V1 as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/UI_TARS | model vlm | 🖥️ Implementing UI-TARS-1.5 as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/MiMo_VL | model vlm | 🎨 Implementing MiMo-VL as a Remote Zoo Model for FiftyOne |
| @harpreetsahota/Kimi_VL_A3B | model vlm | 👀 FiftyOne Remotely Sourced Zoo Model integration for Moonshot AI's Kimi-VL-A3B models enabling object detection, keypoint localization, and image classification with strong GUI and document understanding capabilities. |
| @harpreetsahota/vggt | model 3d | 🎲 Implemeting Meta AI's VGGT as a FiftyOne Remote Zoo Model |
| @harpreetsahota/NVLabs_CRADIOV3 | model embeddings | 📻 Implementing NVLabs C-RADIOv3 Embeddings Model as Remotely Sourced Zoo Model for FiftyOne |
| @harpreetsahota/nemo_retriever_parse_plugin | model ocr | 📜 Implementing NVIDIA NeMo Retriever Parse as a FiftyOne Plugin |
| @harpreetsahota/visual_document_retrieval | model ocr | 📄 A FiftyOne Remotely Sourced Zoo Model integration for LlamaIndex's VDR model enabling natural language search across document images, screenshots, and charts in your datasets. |
| @harpreetsahota/ShowUI | model vlm | 🖥️ Integrating ShowUI into FiftyOne as a Remote Source Zoo Model |
| @harpreetsahota/vitpose | model pose | 🧘🏽 Run ViTPose Models from Hugging Face on your FiftyOne Dataset |
| @harpreetsahota/depth_pro_plugin | model depth | 🥽 Perfom zero-shot metric monocular depth estimation using the Apple Depth Pro model |
| @harpreetsahota/janus_vqa | model vlm | 🐋 Run the Janus Pro Models from Deepseek on your Fiftyone Dataset |
| @harpreetsahota/hiera_video_embeddings | model video | 🎥 Compute embeddings for video using Facebook Hiera Models |
| @segmentsai/segments-voxel51-plugin | annotation | ✏️ Integrate FiftyOne with the Segments.ai annotation tool! |
| @jacobmarks/image_issues | curation | 🌩️ Find common image quality issues in your datasets |
| @jacobmarks/concept_interpolation | curation | 📈 Find images that best interpolate between two text-based extremes! |
| @jacobmarks/text_to_image | model vlm | 🎨 Add synthetic data from prompts with text-to-image models and FiftyOne! |
| @jacobmarks/twilio_automation | data | 📲 Automate data ingestion with Twilio! |
| @wayofsamu/line2d | visualization | 📉 Visualize x,y-Points as a line chart. |
| @jacobmarks/vqa-plugin | model vqa | ❔ Ask (and answer) open-ended visual questions about your images! |
| @jacobmarks/youtube_panel_plugin | visualization | 📺 Play YouTube videos in the FiftyOne App! |
| @jacobmarks/image_deduplication | curation | 🪞 Find exact and approximate duplicates in your dataset! |
| @jacobmarks/keyword_search | search | 🔑 Perform keyword search on a specified field! |
| @jacobmarks/pytesseract_ocr | model ocr | 👓 Run optical character recognition with PyTesseract! |
| @brimoor/pdf-loader | io | 📄 Load your PDF documents into FiftyOne as per-page images |
| @jacobmarks/zero_shot_prediction | model | 🔮 Run zero-shot (open vocabulary) prediction on your data! |
| @jacobmarks/active_learning | annotation | 🏃 Accelerate your data labeling with Active Learning! |
| @jacobmarks/reverse_image_search | curation | ⏪ Find the images in your dataset most similar to an image from filesystem or the internet! |
| @jacobmarks/concept_space_traversal | embeddings | 🌌 Navigate concept space with CLIP, vector search, and FiftyOne! |
| @jacobmarks/audio_retrieval | audio | 🔊 Find the images in your dataset most similar to an audio file! |
| @jacobmarks/semantic_document_search | search | 🔎 Perform semantic search on text in your documents! |
| @allenleetc/model-comparison | evaluation | ⚖️ Compare two object detection models! |
| @ehofesmann/filter_values | search | 🔎 Filter a field of your FiftyOne dataset by one or more values. |
| @jacobmarks/gpt4_vision | model vlm | 🤖 Chat with your images using GPT-4 Vision! |
| @swheaton/anonymize | curation | 🥸 Anonymize/blur images based on a FiftyOne Detections field. |
| @jacobmarks/double_band_filter | search |  Filter on two numeric ranges simultaneously! |
| @danielgural/semantic_video_search | model search |  Semantically search through your video datasets using FiftyOne Brain and Twelve Labs! |
| @jacobmarks/emoji_search | examples | 😏 Semantically search emojis and copy to clipboard! |
| @danielgural/img_to_video | video | 🦋 Bring images to life with image to video! |
| @ehofesmann/edit_label_attributes | annotation | ✏️ Edit attributes of your labels directly in the FiftyOne App! |
| @danielgural/audio_loader | audio visualization | 🎧 Import your audio datasets as spectograms into FiftyOne! |
| @jacobmarks/albumentations_augmentation | data | 🪞 Test out any Albumentations data augmentation transform with FiftyOne! |
| @jacobmarks/image_captioning | model vlm | 🖋️ Caption all your images with state of the art vision language models! |
| @jacobmarks/multimodal_rag | search embeddings | 🦙 Create and test multimodal RAG pipelines with LlamaIndex, Milvus, and FiftyOne! |
| @danielgural/optimal_confidence_threshold | evaluation | 🔍 Find the optimal confidence threshold for your detection models automatically! |
| @danielgural/outlier_detection | curation | ❌ Find those troublesome outliers in your dataset automatically! |
| @danielgural/clustering_algorithms | curation | 🕵️ Find the clusters in your data using some of the best algorithms available! |
| @jacobmarks/clustering | curation | 🍇 Cluster your images using embeddings with FiftyOne and scikit-learn! |
| @mmoollllee/fiftyone-tile | visualization | ⬜ Tile your high resolution images to squares for training small object detection models |
| @mmoollllee/fiftyone-timestamps | curation | 🕒 Compute datetime-related fields (sunrise, dawn, evening, weekday, ...) from your samples' filenames or creation dates |
| @allenleetc/plotly-map-panel | visualization | 🌎 Plotly-based Map Panel with adjustable marker cosmetics! |
| @madave94/multi_annotator_toolkit | annotation | 🧹 Tackle noisy annotation! Find and analyze annotation issues in datasets with multiple annotators per image. |
| @AdonaiVera/fiftyone-vlm-efficient | model vlm curation | 🪄 Improve VLM training data quality with state-of-the-art dataset pruning and quality techniques |
| @AdonaiVera/bddoia-fiftyone | dataset | 🚗 Load and explore the BDDOIA Safe/Unsafe Action dataset via the FiftyOne Zoo |
| @AdonaiVera/fiftyone-agents | vlm evaluation | 🤖 A comprehensive FiftyOne plugin for testing and evaluating multiple vision langugage models with dynamic prompts and built-in evaluation capabilities |
| @AdonaiVera/gemini-vision-plugin | model vlm | 🔮 This plugin integrates Google Gemini's multimodal vision models (e.g., gemini-2.5-flash) into your FiftyOne workflows. Prompt with text and one or more images; receive a text response grounded in visual inputs |
| @allenleetc/sample-inspector | curation | 🔎 Adjust image brightness and contrast and filter semantic masks by class in a sample detail view! |
| @Burhan-Q/fiftyone-vllm | model vlm | 🎯 Run inference using an online vLLM instance for image captioning, classification, object detection, VQA, and OCR. |
| @Burhan-Q/fo-doom | examples | 👾 Play the classic DOOM (1993) shareware game directly within the FiftyOne App. |
| @voxel51/davis-2017 | dataset video | 🚲 Load and explore the DAVIS-2017 video segmentation dataset via the FiftyOne Zoo. |
| @voxel51/mose-v2 | dataset video | ⚽ Load and explore the MOSE complex video object segmentation dataset via the FiftyOne Zoo. |
| @sherpan/torchvision-classifier-finetuner | model training | 🎛️ Fine-tune pretrained torchvision backbones (ResNet-50, EfficientNet-B2, MobileNetV3) on FiftyOne datasets with Classification labels and run inference directly from the App. |
If you haven't already, install FiftyOne:
pip install fiftyoneIn general, you can install all plugin(s) in a GitHub repository by running:
fiftyone plugins download https://github.com/path/to/repoFor instance, to install all plugins in this repository, you can run:
fiftyone plugins download https://github.com/voxel51/fiftyone-pluginsYou can also install a specific plugin using the --plugin-names flag:
fiftyone plugins download \
https://github.com/voxel51/fiftyone-plugins \
--plugin-names <name>💡 Pro tip: Some plugins require additional setup. Click the plugin's link and navigate to the project's README for instructions.
You can use the CLI commands below to manage your downloaded plugins:
# List all plugins you've downloaded
fiftyone plugins list
# List the available operators and panels
fiftyone operators list
# Disable a particular plugin
fiftyone plugins disable <name>
# Enable a particular plugin
fiftyone plugins enable <name>If you plan to develop plugins locally, you can clone the repository and symlink it into your FiftyOne plugins directory like so:
cd /path/to/fiftyone-plugins
ln -s "$(pwd)" "$(fiftyone config plugins_dir)/fiftyone-plugins"Have a plugin you'd like to share with the community? Awesome! 🎉🎉🎉
Just follow these steps to add your plugin to this repository:
- Make sure your plugin repo has a
README.mdfile that describes the plugin and how to install it - Fork this repository
- Add an entry for your plugin to the Community Plugins table above
- Submit a pull request into this repository
You're also welcome to contribute to the plugins that live natively in this repository. Check out the contributions guide for instructions.
If you want join a fast-growing community of engineers, researchers, and practitioners who love computer vision, join the FiftyOne Discord community 🚀🚀🚀
💡 Pro tip: the #plugins channel is a great place to discuss plugins!
If you've made it this far, we'd greatly appreciate if you'd take a moment to check out FiftyOne and give us a star!
FiftyOne is an open source library for building high-quality datasets and computer vision models. It's the engine that powers this project.
Thanks for visiting! 😊