Launch: YOLOv8 Models on Roboflow Universe

Beginning today, pre-trained YOLOv8 models are available on Roboflow Universe. Roboflow Universe is the world's largest source of open-source computer vision datasets, pre-trained models, and computer vision model API's. See our other posts for more information on what's new in YOLOv8 and how to train YOLOv8 on custom data.

Roboflow Universe

This new feature offers any user the ability to:

  1. Test pre-trained YOLOv8 Object Detection models in the app through the Models page.
  2. Deploy pre-trained YOLOv8 Object Detection models through the Roboflow Python package, Hosted Inference API, and browsers and webcams with roboflow.js.

Locating YOLOv8 Models on Roboflow Universe

Datasets with a yolov8n, yolov8s or yolov8 card are those with pre-trained YOLOv8 model weights. You can use the search bar to locate specific model versions.

Face Detection Dataset and Trained YOLOv8s Model

You will also find a note on the Overview page of datasets with pre-trained models indicating the existence of pre-trained YOLOv8n or YOLOv8s model weights.

Construction Site Safety Dataset and Trained YOLOv8 Model

Clicking on the yolov8n, yolov8s or yolov8 card on a dataset will bring up search results for all available models in Roboflow Universe with available YOLOv8n or YOLOv8s model weights.

0:00
/

Testing Pre-Trained YOLOv8 Models

The Model page offers the ability to test models directly in your browser, download datasets, and deploy models to the target device of your choice with available pre-written code. See this construction site safety model as an example:

Construction Site Safety Object Detection Dataset and Pre-Trained Model by Roboflow Universe Projects
717 open source construction images plus a pre-trained Construction Site Safety model and API. Created by Roboflow Universe Projects

Confidence and Overlap settings are available for adjustment based on your preferences. The JSON Response Object from the inference predictions are also available to be inspected, and copied.

Test a model by selecting one of the available images from the dataset's test set; uploading an image or video of your own; uploading a link to a YouTube video; or performing live inference with your webcam.

Deploying Pre-Trained YOLOv8 Models

The Model tab also offers the option to deploy YOLOv8 models with pre-filled code snippets, so you can deploy a model and begin to utilize predictions in your computer vision pipeline immediately.

Use YOLOv8 Models, APIs, and Datasets

YOLOv8 models and APIs are available on Roboflow Universe for use in production applications, automating data labeling via API, sharing model performance with your community, model-assisted labeling in Roboflow Annotate, and more.

As YOLOv8 grows in popularity, we are excited to see model performance across domains and use cases. To contribute a YOLOv8 model, follow our step-by-step YOLOv8 training guide and then upload YOLOv8 weights to your project.