{"id":13642980,"url":"https://github.com/voxel51/voxelgpt","last_synced_at":"2025-06-26T07:01:46.279Z","repository":{"id":173374567,"uuid":"626063658","full_name":"voxel51/voxelgpt","owner":"voxel51","description":"AI assistant that can query visual datasets, search the FiftyOne docs, and answer general computer vision questions","archived":false,"fork":false,"pushed_at":"2024-12-06T21:00:43.000Z","size":304927,"stargazers_count":245,"open_issues_count":8,"forks_count":18,"subscribers_count":22,"default_branch":"main","last_synced_at":"2025-05-31T05:10:36.131Z","etag":null,"topics":["artificial-intelligence","chatgpt","computer-vision","data-science","deep-learning","fiftyone","langchain","llm","machine-learning","openai","python"],"latest_commit_sha":null,"homepage":"https://gpt.fiftyone.ai","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/voxel51.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2023-04-10T18:08:22.000Z","updated_at":"2025-05-05T18:56:55.000Z","dependencies_parsed_at":null,"dependency_job_id":"d96da7a8-aa06-44b0-a3c0-8fd8dc093d47","html_url":"https://github.com/voxel51/voxelgpt","commit_stats":null,"previous_names":["voxel51/voxelgpt"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/voxel51/voxelgpt","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/voxel51%2Fvoxelgpt","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/voxel51%2Fvoxelgpt/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/voxel51%2Fvoxelgpt/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/voxel51%2Fvoxelgpt/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/voxel51","download_url":"https://codeload.github.com/voxel51/voxelgpt/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/voxel51%2Fvoxelgpt/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":262018661,"owners_count":23245614,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["artificial-intelligence","chatgpt","computer-vision","data-science","deep-learning","fiftyone","langchain","llm","machine-learning","openai","python"],"created_at":"2024-08-02T01:01:38.948Z","updated_at":"2025-06-26T07:01:46.241Z","avatar_url":"https://github.com/voxel51.png","language":"Python","funding_links":[],"categories":["Chatbots"],"sub_categories":[],"readme":"# VoxelGPT\n\nWish you could search your images or videos without writing a line of code?\nWant to extract insights from your data by asking in plain English? Now you\ncan! 🎉\n\nVoxelGPT is a [FiftyOne Plugin](https://docs.voxel51.com/plugins/index.html)\nthat combines the power of large language models (LLMs) and large multimodal\nmodels (LMMs) with [FiftyOne](https://github.com/voxel51/fiftyone)'s computer\nvision query language, enabling you to filter, sort, semantically slice and ask\nquestions about your data using natural language. It can even perform\ncomputations on your dataset for you — with approval, of course!\n\n## Live demo\n\n🚀🚀🚀 You can try VoxelGPT live at [gpt.fiftyone.ai](https://gpt.fiftyone.ai)!\n\n## Capabilities\n\nVoxelGPT is capable of handling any of the following types of queries:\n\n-   [Dataset queries](#querying-your-dataset)\n-   [Computation queries](#computational-queries)\n-   [FiftyOne library queries](#fiftyone-library-queries)\n-   [FiftyOne workspace queries](#querying-your-workspace)\n-   [General machine learning queries](#general-machine-learning-queries)\n\nWhen you ask VoxelGPT a question, it will interpret your intent and determine\nwhich type of query you are asking. If VoxelGPT is unsure, it will ask you to\nclarify.\n\n### Querying your dataset\n\nhttps://github.com/voxel51/voxelgpt/assets/12500356/5728b067-defc-4db3-8cda-ad8da3523cf4\n\nVoxelGPT can handle the following types of queries about your dataset:\n\n-   Answer questions about the schema of your dataset, fields, and runs that\n    have been performed\n-   Create a filtered view of your data by constructing and concatenating view\n    stages.\n-   Set the view in the FiftyOne App\n-   Perform aggregations over the entire dataset or a view into the dataset\n\nYou can ask VoxelGPT to search your datasets for you. Here's some examples of\nthings you can ask:\n\n-   Show me 10 random samples\n-   Show me high confidence false positive predictions\n-   Do I have any images with multiple people?\n-   What is the average brightness for my images that contain a cat?\n\nUnder the hood, VoxelGPT interprets your query and translates it into the\ncorresponding\n[dataset view](https://docs.voxel51.com/user_guide/using_views.html). VoxelGPT\nunderstands the schema of your dataset, as well as things like\n[evaluation runs](https://docs.voxel51.com/user_guide/evaluation.html) and\n[similarity indexes](https://docs.voxel51.com/user_guide/brain.html#similarity).\n\nIt can also automatically inspect the contents of your dataset in order to\nretrieve specific entities.\n\n#### Data schema queries\n\nVoxelGPT can answer questions about the schema of your dataset, brain runs,\nevaluation runs, and more. Here are some examples:\n\n-   What fields do I have in my dataset?\n-   Do I have any evaluation runs?\n-   What model did I use to similarity index my dataset?\n\n#### Object detection queries\n\nIf your dataset contains one or more\n[`fo.Detections`](https://docs.voxel51.com/user_guide/using_datasets.html#object-detection)\nfield(s), VoxelGPT can filter or match based on the size (relative and\nabsolute) of bounding boxes, and on the number of detections.\n\n-   Restrict the view to ground truth detections larger than half of the image\n    area\n-   Show me all of the predictions \u003c $96^2$ pixels\n-   What is the average number of person detections I have per image?\n\n#### Geolocation queries\n\nIf your dataset has a\n[`GeoLocation`](https://docs.voxel51.com/user_guide/using_datasets.html#geolocation)\nfield, you can run geographic queries on your dataset. VoxelGPT can perform\ngeocoding to go from location name (or textual description) to a `(lon, lat)`\npair, or a list of `(lon, lat)` points defining a boundary region. Here are a\nfew examples:\n\n-   Sort by proximity to the Statue of Liberty\n-   Show me samples within 400m of Grand Central\n-   Filter for images of Paris\n-   How many images do I have that were taken in Hell's Kitchen?\n\n#### Temporal queries\n\nIf your dataset has a\n[`Date` or `DateTime` field](https://docs.voxel51.com/user_guide/basics.html#fields),\nVoxelGPT can perform temporal queries such as:\n\n-   Filter for pictures taken on a Tuesday\n-   How many images were added after June 01, 2023?\n-   Show me samples with `event` field reading a time of day between 8pm and\n    11pm\n\n#### Aggregations\n\nVoxelGPT has access to Aggregation stages in FiftyOne, so it can perform\naggregations like `count`, `mean`, `sum`, `std`, `min`, `max`, `values`, and\n`distinct` for a field or expression over the entire dataset or a view into the\ndataset. Here are some examples:\n\n-   What is the average brightness of my images?\n-   How many images do I have with a `cat` label?\n-   What is the standard deviation of the `confidence` field in my predictions?\n\n### Computational queries\n\nVoxelGPT can perform computations on your dataset, such as:\n\n-   Brightness: assign a brightness score to each sample in the dataset, using\n    FiftyOne's\n    [Image Quality Issues plugin](https://github.com/jacobmarks/image-quality-issues)\n-   Entropy: quantify the amount of information in each sample in the dataset,\n    using FiftyOne's\n    [Image Quality Issues plugin](https://github.com/jacobmarks/image-quality-issues)\n-   Uniqueness: assign a uniqueness score to each sample in the dataset, using\n    the [FiftyOne Brain](https://voxel51.com/fiftyone/workflows/uniqueness/)\n-   Duplicates: identify and remove duplicate samples in the dataset, using the\n    [FiftyOne Brain](https://docs.voxel51.com/api/fiftyone.brain.html?highlight=duplicate#fiftyone.brain.compute_exact_duplicates)\n-   Similarity: generate a vector similarity index on the dataset, which can be\n    used to compare samples in the dataset, using the\n    [FiftyOne Brain](https://docs.voxel51.com/user_guide/brain.html#similarity)\n-   Dimensionality reduction: reduce the dimensionality of feature vectors for\n    each sample, using the\n    [FiftyOne Brain](https://docs.voxel51.com/user_guide/brain.html#visualizing-embeddings)\n    using UMAP, PCA, or t-SNE, so that they can be visualized in 2D or 3D\n-   Clustering: cluster samples in the dataset using KMeans, DBSCAN, and other\n    clustering algorithms, using FiftyOne's\n    [Clustering plugin](https://github.com/jacobmarks/clustering-plugin)\n\nHere's some examples of computational queries you can ask VoxelGPT:\n\n-   Compute the brightness of images across my dataset\n-   Score the uniqueness of each image in my dataset\n-   Generate a similarity index for my dataset\n-   Cluster my dataset using KMeans\n-   Help me visualize my dataset in 2D using UMAP\n\n💡 If you do not want to allow VoxelGPT to run computations, set the\nenvironment variable:\n\n```shell\nexport VOXELGPT_ALLOW_COMPUTATIONS=false\n```\n\nYou can also set the minimum dataset size at which VoxelGPT needs to ask for\npermission to run computations:\n\n```shell\nexport VOXELGPT_APPROVAL_THRESHOLD=1000\n```\n\nThe default value is 100 samples.\n\n### FiftyOne library queries\n\nVoxelGPT is not only a pair programmer; it is also an educational tool.\nVoxelGPT has access to the entire [FiftyOne docs](https://docs.voxel51.com), as\nwell as all of the blog posts on the [Voxel51 Blog](https://voxel51.com/blog/),\nand transcripts from videos on the\n[Voxel51 YouTube channel](https://www.youtube.com/channel/UC9GWqiVDwPdQrW70_v4VtlQ).\nIt can use all of these resources to answer FiftyOne-related questions.\n\nHere's some examples of documentation queries you can ask VoxelGPT:\n\n-   How do I load a dataset from the FiftyOne Zoo?\n-   What does the match() stage do?\n-   Can I export my dataset in COCO format?\n-   Does FiftyOne have any plugins for active learning?\n\nVoxelGPT will provide links to the most helpful resources across Voxel51's\ndocs, blog, and YouTube channel. For YouTube videos, the links will point\ndirectly to the most relevant timestamp!\n\n### Querying your workspace\n\nVoxelGPT can answer questions about the environment in which you are running\nFiftyOne, including:\n\n-   Other datasets you have downloaded\n-   Plugins you have installed, and operators within those plugins\n-   Your FiftyOne config\n-   Your FiftyOne App config\n\nHere's some examples of workspace queries you can ask VoxelGPT:\n\n-   Do I have any COCO datasets?\n-   Do I have any plugins for identifying issues in my data?\n-   What is my operator timeout set to?\n\n### General machine learning queries\n\nhttps://github.com/voxel51/voxelgpt/assets/12500356/294b53f8-9398-4e6a-b923-56c7a9684f1d\n\nFinally, VoxelGPT can answer general questions about computer vision, machine\nlearning, and data science. It can help you to understand basic concepts and\nlearn how to overcome data quality issues.\n\nHere's some examples of machine learning queries you can ask VoxelGPT:\n\n-   What is the difference between precision and recall?\n-   How can I detect faces in my images?\n-   What are some ways I can reduce redundancy in my dataset?\n\n## Installation\n\nIf you haven't already, install\n[FiftyOne](https://github.com/voxel51/fiftyone):\n\n```shell\npip install fiftyone\n```\n\nYou'll also need to provide an OpenAI API key\n([create one](https://platform.openai.com/account/api-keys)):\n\n```shell\nexport OPENAI_API_KEY=XXXXXXXX\n```\n\nFor use with your private Azure deployment, see\n[here](#using-azure-openai-deployment)\n\n### App-only use\n\nIf you only want to use VoxelGPT in the\n[FiftyOne App](https://docs.voxel51.com/user_guide/app.html), then you can\nsimply [install it as a plugin](https://docs.voxel51.com/plugins/index.html):\n\n```shell\nfiftyone plugins download https://github.com/voxel51/voxelgpt\nfiftyone plugins requirements @voxel51/voxelgpt --install\n```\n\n### Local use/development\n\nIf you want to directly use the `voxelgpt` module or develop the project\nlocally, then you'll want to clone repository:\n\n```shell\ngit clone https://github.com/voxel51/voxelgpt\ncd voxelgpt\n```\n\ninstall the requirements:\n\n```shell\npip install -r requirements.txt\n```\n\nand make the plugin available for use in the FiftyOne App by symlinking it into\nyour plugins directory:\n\n```shell\n# Symlinks your clone of voxelgpt into your FiftyOne plugins directory\nln -s \"$(pwd)\" \"$(fiftyone config plugins_dir)/voxelgpt\"\n```\n\n### FiftyOne Teams\n\nWant to add VoxelGPT to your\n[FiftyOne Teams](https://voxel51.com/fiftyone-teams) deployment? You can!\n[Instructions here](FIFTYONE_TEAMS.md).\n\n### Using an Azure OpenAI deployment\n\nYou can use VoxelGPT with your private Azure deployment by setting the\nfollowing environment variables:\n\n```shell\nexport OPENAI_API_TYPE=azure\nexport AZURE_OPENAI_ENDPOINT=\u003cazure_endpoint\u003e\nexport AZURE_OPENAI_KEY=\u003cazure_api_key\u003e\n\nexport AZURE_OPENAI_GPT35_DEPLOYMENT_NAME=\u003cgpt35-deployment-name\u003e\nexport AZURE_OPENAI_GPT4O_DEPLOYMENT_NAME=\u003cgpt4o-deployment-name\u003e\nexport AZURE_OPENAI_TEXT_EMBEDDING_3_LARGE_DEPLOYMENT_NAME=\u003cembedding-deployment-name\u003e\n```\n\nIf any of the first three environment variables is not set, VoxelGPT will\ndefault to using the OpenAI API. For the last three environment variables, if\nany of them is not set of the resource is not found, VoxelGPT will default to\nusing the OpenAI API for that specific model.\n\n## Using VoxelGPT in the App\n\nYou can use VoxelGPT in the FiftyOne App by loading any dataset:\n\n```py\nimport fiftyone as fo\nimport fiftyone.zoo as foz\n\ndataset = foz.load_zoo_dataset(\"quickstart\")\nsession = fo.launch_app(dataset)\n```\n\nand then either:\n\n-   Clicking on the OpenAI icon above the grid\n-   Pressing the `+` icon next to the Samples tab and choosing VoxelGPT\n\nhttps://github.com/voxel51/voxelgpt/assets/12500356/fbaccb6f-dc7f-43d7-9da3-adb4062c398b\n\nFor example, try asking the following questions:\n\n-   What are some popular model architectures for computer vision?\n-   How do I search for specific objects in my dataset?\n-   Show me predicted airplanes\n\n**Pro tip:** use the [`now` keyword](#keywords) to incorporate your previous\nprompts as context for your next query!\n\nYou can also run VoxelGPT as an\n[operator](https://docs.voxel51.com/plugins/index.html#fiftyone-operators) by\npressing the `~` key on your keyboard and selecting `Ask VoxelGPT` from the\nlist. This will open up a small modal where you can type in your query.\n\n## Using VoxelGPT in Python\n\nIf you've [installed locally](local-use/development), you can also directly\ninteract with VoxelGPT via Python.\n\n### Interactive sessions\n\nYou can use `ask_voxelgpt_interactive()` to launch an interactive session where\nyou can converse with VoxelGPT via `input()` prompts. Navigate to the directory\nwhere VoxelGPT is located and run:\n\n```py\nimport fiftyone as fo\nimport fiftyone.zoo as foz\n\nfrom voxelgpt import ask_voxelgpt_interactive\n\ndataset = foz.load_zoo_dataset(\"quickstart\")\nsession = fo.launch_app(dataset)\n\nask_voxelgpt_interactive(dataset, session=session)\n```\n\n**Pro tip:** use the [`now` keyword](#keywords) to incorporate your previous\nprompts as context for your next query.\n\nAs usual, you can prompt VoxelGPT with any combination of dataset,\ndocumentation, and general computer vision queries. For example, a conversation\nmight look like:\n\n```\nYou:\n    what is the difference between precision and recall?\n\nVoxelGPT:\n    Precision and recall are two important metrics used to evaluate the\n    performance of a machine learning model in classification tasks. Precision\n    measures the proportion of true positive predictions among all positive\n    predictions made by the model. Recall, on the other hand, measures the\n    proportion of true positive predictions among all actual positive instances\n    in the dataset.\n\n    In other words, precision is the ability of the model to correctly identify\n    positive instances, while recall is the ability of the model to identify\n    all positive instances. A high precision means that the model is making\n    very few false positive predictions, while a high recall means that the\n    model is able to identify most of the positive instances in the dataset.\n\n    It is important to balance precision and recall when evaluating a model's\n    performance. A model with high precision but low recall may be missing\n    important positive instances, while a model with high recall but low\n    precision may be making too many false positive predictions.\n\n    In the context of the FiftyOne library, precision and recall can be\n    computed using the `fiftyone.utils.metrics` module, which provides\n    functions for computing various classification metrics. For example, the\n    `fiftyone.utils.metrics.binary_classification` function can be used to\n    compute precision and recall for binary classification tasks.\n\nYou:\n    does FiftyOne integrate with Label Studio?\n\nVoxelGPT:\n    Yes, FiftyOne integrates with Label Studio. The integration allows you to\n    easily upload your data directly from FiftyOne to Label Studio for labeling.\n    You can get started with Label Studio through a simple pip install to get a\n    local server up and running. FiftyOne provides simple setup instructions\n    that you can use to specify the necessary account credentials and server\n    endpoint to use. FiftyOne provides an API to create projects, upload data,\n    define label schemas, and download annotations using Label Studio, all\n    programmatically in Python. All of the following label types are supported\n    for image datasets: Classification, Detections, Instance segmentations,\n    Polygons and polylines, Keypoints, Scalar fields, and Semantic segmentation.\n\n    Sources:\n    - https://docs.voxel51.com/integrations/index.html\n    - https://docs.voxel51.com/integrations/labelstudio.html\n\nYou:\n    show me predicted airplanes\n\nVoxelGPT:\n    Identified potential algorithms: evaluation\n    No evaluation runs found.\n    Identified potential fields: predictions\n    Identified potential label classes: {'predictions': ['airplane']}\n    Identified potential view stages: ['match', 'filter_labels', 'match_labels', 'exclude_labels', 'filter_field']\n    Okay, I'm going to load dataset.filter_labels(\"predictions\",F(\"label\") == \"airplane\")\n\nYou:\n    now only show me the first 10 samples\n\nVoxelGPT:\n    Identified potential algorithms: evaluation\n    No evaluation runs found.\n    Identified potential fields: predictions\n    Identified potential label classes: {'predictions': ['airplane']}\n    Identified potential view stages: ['match', 'limit', 'limit_labels', 'skip', 'sort_by']\n    Okay, I'm going to load dataset.match(F(\"predictions.detections.label\").contains(\"airplane\")).limit(10)\n\nYou: exit\n```\n\nIn interactive mode, VoxelGPT automatically loads any views it creates in the\nApp, and you can access them via your\n[session](https://docs.voxel51.com/user_guide/app.html#sessions) object:\n\n```py\nprint(session.view.count(\"predictions.detections\"))\n```\n\n### Single queries\n\nYou can also use `ask_voxelgpt()` to prompt VoxelGPT with individual queries:\n\n```py\nfrom voxelgpt import ask_voxelgpt\n\nask_voxelgpt(\"Does FiftyOne integrate with CVAT?\")\n```\n\n```\nYes, FiftyOne integrates with CVAT, which is an open-source image and video\nannotation tool. You can upload your data directly from FiftyOne to CVAT to add or\nedit labels. You can use CVAT either through the hosted server at app.cvat.ai or\nthrough a self-hosted server. In either case, FiftyOne provides simple setup\ninstructions that you can use to specify the necessary account credentials and\nserver endpoint to use. The tight integration between FiftyOne and CVAT allows\nyou to curate and explore datasets in FiftyOne and then send off samples or\nexisting labels for annotation in CVAT with just one line of code. To use CVAT,\nyou must create an account on a CVAT server. By default, FiftyOne uses app.cvat.ai.\nIf you haven’t already, go to app.cvat.ai and create an account now. Another option\nis to set up CVAT locally and then configure FiftyOne to use your self-hosted server.\nA primary benefit of setting up CVAT locally is that you are limited to 10 tasks and\n500MB of data with app.cvat.ai.\n\nSources:\n- https://docs.voxel51.com/integrations/cvat.html#examples\n- https://docs.voxel51.com/tutorials/cvat_annotation.html#Annotating-Datasets-with-CVAT\n- https://docs.voxel51.com/tutorials/cvat_annotation.html#Setup\n- https://docs.voxel51.com/integrations/index.html#fiftyone-integrations\n```\n\nWhen VoxelGPT creates a view in response to your query, it is returned:\n\n```py\nimport fiftyone as fo\nimport fiftyone.zoo as foz\n\ndataset = foz.load_zoo_dataset(\"quickstart\")\n\nview = ask_voxelgpt(\"show me 10 random samples\", dataset)\n```\n\n```\nIdentified potential view stages: ['match', 'limit', 'skip', 'take', 'sort_by']\nOkay, I'm going to load dataset.take(10)\n```\n\n## Keywords\n\nVoxelGPT is trained to recognize certain keywords that help it understand your\nintent:\n\n| Keyword                 | Meaning                                                                                                                                                                                                                                                      |\n| ----------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |\n| `show`/`display`        | Tells VoxelGPT that you want it to query your dataset and display the results                                                                                                                                                                                |\n| `docs`/`how`/`FiftyOne` | Tells VoxelGPT that you want it to query the FiftyOne docs.                                                                                                                                                                                                  |\n| `now`                   | Use your chat history as context to interpret your next query. For example, if you ask \"show me images with people\" and then ask \"now show me the 10 most unique ones\", VoxelGPT will understand that you want to show the 10 most unique images with people |\n| `help`                  | Prints a help message with usage instructions                                                                                                                                                                                                                |\n| `reset`                 | Resets the conversation history                                                                                                                                                                                                                              |\n| `exit`                  | Exits interactive Python sessions                                                                                                                                                                                                                            |\n\n## Contributing\n\nContributions are welcome! Check out the [contributions guide](CONTRIBUTING.md)\nfor instructions.\n\n## How does it work?\n\nVoxelGPT uses:\n\n-   OpenAI's\n    [GPT-3.5-Turbo](https://platform.openai.com/docs/models/gpt-3-5-turbo) and\n    [GPT-4o](https://platform.openai.com/docs/models/gpt-4o) to generate\n    textual answers and code\n-   OpenAI's\n    [text-embedding-3-large model](https://platform.openai.com/docs/guides/embeddings/embedding-models)\n    to embed input text prompts\n-   [LangChain](https://github.com/hwchase17/langchain) provides the connective\n    tissue for the application\n-   FiftyOne's [plugin framework](https://docs.voxel51.com/plugins/index.html)\n    to provide the interactive panel in the\n    [FiftyOne App](https://docs.voxel51.com/user_guide/app.html)\n\n## Limitations\n\n### Media types\n\nVoxelGPT provides limited support for videos, grouped datasets, and 3D media.\nBasic filtering, querying, and aggregations will still work, but don't expect\ndeep insights into 3D data.\n\n### Examples\n\nThis implementation is based on a limited set of examples, so it may not\ngeneralize well to all datasets. The more specific your query, the better the\nresults will be. If you find that the results are not what you expect, please\nlet us know!\n\n## About FiftyOne\n\nIf you've made it this far, we'd greatly appreciate if you'd take a moment to\ncheck out [FiftyOne](https://github.com/voxel51/fiftyone) and give us a star!\n\nFiftyOne is an open source library for building high-quality datasets and\ncomputer vision models. It's the engine that powers this project.\n\nThanks for visiting! 😊\n\n## Join the Community\n\nIf you want join a fast-growing community of engineers, researchers, and\npractitioners who love visual AI, join the\n[FiftyOne Slack community](https://slack.voxel51.com/)! 🚀🚀🚀\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fvoxel51%2Fvoxelgpt","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fvoxel51%2Fvoxelgpt","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fvoxel51%2Fvoxelgpt/lists"}