Get Started

Explore Datasets, Models, and Projects

The Ultralytics Platform Explore page lists public content from the community. Find datasets to train on, trained models by what they detect, and projects for inspiration and learning. The Explore page is accessible to everyone — even without signing in.

Ultralytics Platform Explore Datasets Tab Cards View

Anonymous Access

The Explore page works without signing in. Signed-out visitors get Datasets and Models sections in the sidebar listing official @ultralytics content, in place of the Annotate, Train, and Deployments sections that hold your own work. Public models can even be tested in the browser from their Predict tab. To clone content or create your own, you'll need to sign up.

Overview#

The Explore page features four tabs:

  • Public Datasets: Community training data with image previews
  • Search Images: Images from public datasets, and from your own and team datasets when signed in, found by what they show
  • Public Models: Trained models found by what they detect, each with its training images and headline metrics
  • Public Projects: Complete experiments containing trained models

Only datasets that contain at least one image and projects that contain at least one model appear on Explore, so empty placeholders never surface in results.

Browse Content#

Tabs#

The Explore page uses a tabbed interface with Datasets, Images, Models, and Projects tabs. Each tab has its own search; the Datasets, Models, and Projects tabs also have sort and view mode controls.

TabDescription
DatasetsLabeled image collections for training (default)
ImagesImages found by what they show
ModelsTrained models found by what they detect
ProjectsOrganized model collections with training results

Each tab has its own search bar. Results update as you type and the query is kept in the URL as ?q=, so a search is easy to bookmark or share.

Ultralytics Platform Explore Datasets Tab With Search

TabMatched Fields
DatasetsName, description, owner username, tags, and class names
ImagesWhat the image shows
ModelsModel name; the project's name, description, owner username, and tags; and what its training images show
ProjectsName, description, owner username, and tags

On the Datasets tab, datasets that match the fields above come first, followed by datasets whose images match the search, and their cards show the matching images. Searching an official dataset's name or address (such as HIT-UAV or hit uav) puts the official @ultralytics dataset above copies that share its name. Before you search, the Images tab lists one image from each of the newest public datasets. Search it by what images show with semantic search: describe a scene such as cats on a sofa. Click any image to open a side panel with its annotations, source dataset, class counts, and license, followed by up to 24 similar images, including ones from the same dataset, when you are signed in; click one to open it in the panel. On narrow screens such as phones, the image opens full size instead.

To search by a photo instead, click the search-by-image icon in the Datasets, Images, or Models search bar, or drop an image onto the page. iPhone HEIC photos work too: the browser converts them to JPEG before searching. The Images tab shows images like it, each with its similarity score, the Datasets tab shows public datasets containing similar images, with those images on their cards, and the Models tab shows models trained on similar images. Filters, and sorting on the Datasets and Models tabs, still apply, and switching between these three tabs keeps the photo. The photo is searched, not saved, and stays out of the URL; the search bar shows it until you click Close, type a search, or open the Projects tab.

Find Models by What They Detect#

On the Models tab, describe what you need to detect, such as hard hat or aerial drones, to find public trained models even when no model or project name contains your words.

Ultralytics Platform Explore Models Tab Search Results

Results are ranked in three tiers:

  1. Official @ultralytics models whose name or project matches, such as YOLO26 models for yolo26
  2. Models whose name or project matches the search
  3. Models trained on public datasets whose images match the search, best match first

To keep repeated runs and copies from filling the page, the tab lists one model per training dataset (or per project for models without one), while official @ultralytics models each keep their own card. It leaves out cloned models and models that haven't finished training; the training dataset's Models tab lists the other models trained on it. Before you search, the tab lists models trained on public datasets, newest first. Models without a public training dataset, such as uploaded .pt files, are found by their model or project name only.

Click a card to open the model, then test it from its Predict tab, clone it and deploy your copy, or download it.

Sort and Filter#

The dropdown in the toolbar holds sort options on the Datasets, Models, and Projects tabs and filters on all four tabs. The Models tab sorts by Relevance or Created only:

Sort OptionDescription
RelevanceDatasets and Models tab default: best search matches first, newest first without a search
StarsCommunity star count (descending)
CreatedCreation date, newest or oldest first; newest is the Projects tab default
NameAlphabetical, ascending or descending
Images / ModelsImage count (datasets) or model count (projects), most or fewest first
FilterTabsAvailabilityDescription
TaskDatasets, Images, ModelsEveryoneShow only content for the selected tasks (detect, segment, semantic, depth, classify, pose, obb)
My starredDatasets, Images, ProjectsSigned in onlyShow only datasets or projects you have starred, or images from those datasets
Image countDatasetsEveryoneShow only datasets with Under 1K, 1K – 10K, 10K – 100K, or 100K+ images
LicenseDatasets, Images, ProjectsEveryoneShow only datasets (or images from them) or projects published under the selected licenses

Sorting and filtering are also stored in the URL (dsort, msort, psort, task, starred, imageCount, license, and plicense), so the browser back button restores the previous result list.

Sorting by Stars

The Stars sort shows only datasets with at least one star, so unrated datasets are hidden while it is active.

View Modes#

On the Datasets and Projects tabs, toggle between three view modes for browsing; the Models tab offers Cards and Compact:

ModeDescription
CardsGrid of preview cards with thumbnails
CompactSmaller cards in a responsive grid (2-3 columns)
TableTable with sortable columns

All three views load more results as you scroll. Table view moves sorting from the dropdown to sortable column headers and keeps the filters in the dropdown.

Content Cards#

Each item displays:

Ultralytics Platform Explore Projects Tab With Project Cards

ElementDescription
IconProject icon with custom color, letter, or cover image
NameProject title
CreatorAuthor avatar and username
DescriptionShort project description
Star CountNumber of community stars (shown when above zero)
Model CountNumber of models in the project
SizeTotal storage used by the project's model files
UpdatedDate the project last changed
Model TagsNames of the first few models in the project
TagsProject tags

Use Public Content#

Clone Dataset#

Use a public dataset for your training:

  1. Click on the dataset to open its detail page
  2. Click Clone Dataset
  3. Review the destination workspace, name, visibility, and license
  4. Click Clone Dataset to confirm

Ultralytics Platform Explore Clone Dataset

Cloned Dataset Properties
  • The clone dialog lets you review the destination, name, visibility, and license
  • Public datasets stay public by default unless the workspace's default dataset visibility is Private
  • A copyleft source license is carried over and locked, so the clone keeps the terms it was published under
  • For a new dataset, the clone dialog blocks a URL already used in the target workspace until you choose another
  • You can modify classes, annotations, and splits
  • Changes don't affect the original dataset
  • Image files are reused rather than copied, but the clone still counts toward the destination workspace's storage quota

See Datasets for managing and annotating your cloned dataset.

Download Model#

Download a public model:

  1. Open the model from the Models tab or from its project
  2. Click the download icon to download its PyTorch .pt weights
  3. Open the Export tab to download any completed exports that are available

You can also use the model for inference or as a starting point for fine-tuning:

# Use a downloaded model for inference
yolo predict model=path/to/downloaded-model.pt source=image.jpg

# Fine-tune on your own dataset
yolo train model=path/to/downloaded-model.pt data=my-dataset.yaml epochs=50

Clone Model#

Clone a public model to one of your projects:

  1. Open the model from the Models tab or from its project
  2. Click Clone Model
  3. Select a Project, or choose New Project to create one
  4. Optionally, rename the model
  5. Review the source model summary, then click Clone Model to confirm

Ultralytics Platform Explore Clone Model Dialog

Clone vs Download

Clone copies the model to a project on the Platform for further training or deployment. Download saves the model file to your local machine.

Clone Project#

Copy a public project to your workspace:

  1. Click on the project to open its detail page
  2. Click Clone Project
  3. Review the destination, name, visibility, and license, then click Clone Project

Ultralytics Platform Explore Clone Project

The cloned project includes completed models that have model files. Deployments and exports are not copied. The models count toward your workspace's model limit. Its default visibility follows the same workspace rules as dataset cloning.

See Projects for organizing models in your project.

Official Ultralytics Content#

Official @ultralytics projects are pinned to the top of the Explore Projects tab, and their models lead matching Models tab searches: YOLO26, YOLO11, YOLOv8, and YOLOv5. Open a project to see its current models and supported tasks.

Official datasets include benchmark datasets like coco8 (8-image COCO subset), VOC, african-wildlife, dota8, and other commonly used computer vision datasets.

Quick Start with Official Models

The fastest way to get started is to fine-tune an official pretrained model on your own dataset — no cloning required:

  1. Upload your dataset in supported formats
  2. Click New Model on the dataset page
  3. Choose a YOLO26 model from the Official tab of the base model selector and start training

User Profiles#

Click on a creator's username to view their public profile at platform.ultralytics.com/{username}. Public profiles show:

Ultralytics Platform User Profile Public Content

SectionContent
IdentityAvatar, display name, @username, and plan badge
BioUser description and company
LinksSocial profiles
FollowersFollower count
DatasetsPublic datasets (default tab)
ProjectsPublic projects with models

The Datasets and Projects tabs use the same search, sort, and view controls as the Explore page, without the Relevance sort.

Follow Users#

Click the Follow button on any user's profile to follow or unfollow them. Follower counts are displayed on public profiles. The Share button next to it copies a link or opens a pre-filled social post.

Make Your Content Public#

Make your work available to the community. Public content is visible to everyone, including users who aren't signed in. Explore lists public datasets that contain at least one image and public projects that contain at least one model, so an empty dataset or a project with no models stays off the Explore page until you add content to it. Models follow the visibility of the project that holds them and have no separate badge; a public project's trained models, except clones, are listed on the Models tab.

Make Dataset Public#

  1. Go to your dataset
  2. Click the Private visibility badge in the top navigation bar
  3. Review the warning and click Make public

Make Project Public#

  1. Go to your project
  2. Click the Private visibility badge in the top navigation bar
  3. Review the warning and click Make public
Quality Content

Before making content public:

  • Add a clear, descriptive name and description
  • Define class names in the dataset settings
  • Verify data quality and annotation accuracy
  • Test model performance and include training metrics
Public Content Visibility

Public content is visible to everyone on the internet, including anonymous users. Make sure your dataset doesn't contain sensitive, private, or copyrighted data before making it public. You can change visibility back to private at any time.

Guidelines#

When contributing public content:

Do#

  • Provide useful, high-quality content
  • Write clear descriptions
  • Include relevant metadata
  • Respond to questions
  • Credit data sources

Don't#

  • Upload sensitive/private data
  • Violate copyrights
  • Upload inappropriate content
  • Spam low-quality content
  • Misrepresent performance

Share Content#

Click the Share button on any public profile, project, model, or dataset. On desktop it opens a menu with Share on X, Share on LinkedIn, Copy link, and — for embeddable resources — Copy embed code. The social options open a pre-filled post. On mobile it hands off to your device's native share sheet.

Embed Widgets#

Public content can be embedded in external websites using embed URLs:

ContentEmbed URL Pattern
Projectplatform.ultralytics.com/embed/{username}/{project}
Modelplatform.ultralytics.com/embed/{username}/{project}/{model}
Datasetplatform.ultralytics.com/embed/{username}/datasets/{slug}

Use these URLs in an <iframe> to embed public preview cards for projects, models, or datasets. Each card links to the full resource on Platform. Copy embed code in the Share menu produces the <iframe> snippet for you.

Public Content URLs#

Public content on the platform uses clean, shareable URLs:

ContentURL PatternExample
Profileplatform.ultralytics.com/{username}platform.ultralytics.com/ultralytics
Datasetsplatform.ultralytics.com/{username}/datasetsplatform.ultralytics.com/ultralytics/datasets
Datasetplatform.ultralytics.com/{username}/datasets/{slug}platform.ultralytics.com/ultralytics/datasets/coco8
Projectplatform.ultralytics.com/{username}/{project}platform.ultralytics.com/ultralytics/yolo26
Modelplatform.ultralytics.com/{username}/{project}/{model}platform.ultralytics.com/ultralytics/yolo26/yolo26n
Shareable Links

You can share any public content URL directly. Recipients can view the content without signing in. To clone or download, they'll need an account.

FAQ#

  • Check the license shown on the individual project, model, or dataset. Usage rights depend on that resource's license; contact the creator when the license does not grant the rights you need.

  • To report inappropriate content:

    1. Navigate to the public page containing the content (project or dataset), if accessible
    2. Open the Help page from the sidebar
    3. Select General as the feedback type
    4. Describe the content and the issue, including a link to the page
    5. Submit the report

    If the content is no longer accessible, use the Help page from any page and include as much detail as possible (URL, username, or description).

  • Yes, you can change visibility anytime:

    1. Open the project or dataset
    2. Click the Public visibility badge in the top navigation bar
    3. The resource changes to private immediately

    Existing clones are not affected.

Comments

Ask a question, share your results, or help someone else.