Join us as we take a closer look at how you can use Ultralytics HUB to train and deploy the new Ultralytics YOLO11 models. We'll walk you through the process step-by-step.
Ultralytics YOLO11 is the new state-of-the-art computer vision model designed for tasks like object detection, image classification, and instance segmentation. It’s faster, more accurate, and more efficient than previous versions of YOLO (You Only Look Once) models. YOLO11 can be used for a variety of real-time computer vision applications. Best of all, getting started with Ultralytics YOLO11 is just as simple and straightforward as all other Ultralytics YOLO models.
We previously discussed YOLO11’s new features and improvements and touched on accessing the model through the Ultralytics Python package or Ultralytics HUB. In this guide, we’ll walk you through how to use Ultralytics HUB step-by-step to train and deploy Ultralytics YOLO11 easily.
Ultralytics HUB is Ultralytics’ no-code, user-friendly platform designed to streamline the entire process from training to deploying YOLO models, including the newly launched Ultralytics YOLO11 models. Whether you’re an AI expert or new to computer vision, the HUB provides an intuitive interface that allows you to upload datasets, select pre-trained models, and fine-tune them for your specific needs. With just a few clicks, you can train models for real-time applications in industries ranging from manufacturing to agriculture. HUB focuses on making advanced AI accessible without the need for extensive coding.
Ultralytics HUB has different plan options, with a free tier for basic access and a Pro plan offering additional capabilities like cloud training, team collaboration, and increased usage limits. Here’s a quick glance at some of the key features offered by Ultralytics HUB:
HUB also integrates with various platforms, and users can export trained models to various formats such as ONNX, TensorFlow, and CoreML, making deployment across multiple platforms seamless. Essentially, Ultralytics HUB simplifies complex AI tasks, from dataset handling to real-time model deployment, all within one comprehensive tool.
To run inferences on Ultralytics HUB using YOLO11, simply navigate to the "Models" section and choose the YOLO11 model you’re interested in. Then, you can click on "Preview" to try the model by uploading any image.
This feature of HUB makes it possible for anyone, regardless of their experience level, to test model predictions with YOLO11 and see how it performs. It's a user-friendly way to get hands-on with Ultralytics YOLO11 for free.
After creating an account, you can dive right into training by accessing the dashboard. From there, you can manage your projects, upload datasets, and start training your YOLO11 models with ease. The platform is designed to keep the process quick and as hassle-free as possible.
Once you are logged in, you can click on "Datasets" from the menu on the left to explore a range of pre-existing datasets available on Ultralytics HUB. These datasets cater to various tasks, such as oriented bounding boxes (OBB) object detection and pose estimation. For example, you can use COCO128 for object detection with 80 classes or Fashion-MNIST for image classification. These datasets are readily available and optimized for training YOLO models.
If you'd like to work with your own data, you can upload custom datasets. When doing so, make sure that your dataset follows the YOLO structure, including a properly formatted YAML file in the root directory, and that it is zipped.
Once your dataset is ready, you can click on the "Upload Dataset" button, select the task type, and upload the ZIP file. After uploading, Ultralytics HUB automatically validates your dataset, and you can immediately begin training YOLO models. You can also manage and view your dataset details, such as image splits (train, validation, test), and analyze data to ensure it's ready for model training.
To begin training a YOLO11 model using Ultralytics HUB’s Cloud Training feature, you’ll need to upgrade to the Pro plan. As a Pro user, GPU resources are available to you for faster and more efficient training. Once you’re upgraded, access the “Models” section, select your desired YOLO11 model variation, and configure the training settings.
You can choose the number of epochs (which define how many times the model will pass through the dataset) or set a specific duration for timed training. Before the model training begins, Ultralytics HUB will initialize a dedicated GPU instance to ensure optimized performance. Depending on demand, the initialization may take some time, but no charges will be applied to your account during this process.
After finalizing your settings, click "Start Training" to launch the session. Throughout the training, you can monitor progress in real-time through a dashboard. It gives you the ability to pause, stop, or resume training as needed. If your account balance runs low during epoch-based training, the session will pause, allowing you to top up your balance before resuming. The platform automatically saves checkpoints, meaning you can pick up from where you left off.
At the end of the training, you can check all costs through the billing tab, where you’ll find detailed cost reports that make it easy to track expenses and manage your training efficiently.
When deploying your custom-trained YOLO11 model with Ultralytics HUB, there are two main options: the Shared Inference API and the Dedicated Inference API. To use the deployed model, you can make inference requests to the API using either Python or cURL, depending on your setup. The general process involves sending an image file along with relevant parameters (like image size and confidence thresholds) to the API. Ultralytics HUB will return the predictions in a simple JSON format, which you can process further.
The Shared Inference API is a cost-effective solution for users on the free tier and provides 100 calls per hour and up to 1000 calls monthly. It eliminates the need for a local environment and supports quick deployment directly from the Ultralytics HUB.
The Dedicated Inference API, available to Pro users, is more suitable for larger-scale deployments or real-time applications. It provides a single-click deployment in a dedicated cloud environment powered by Google Cloud Run. This option is optimized for high-performance applications, ensuring sub-100ms latency and global coverage across 38 regions for real-time processing. It also supports enhanced security features, making it suitable for industries with stringent data protection requirements.
Once you’ve chosen between the Shared or Dedicated Inference API for deploying your YOLO11 model, the next steps are simple and efficient. You can open the “Deploy” tab within your model's page on Ultralytics HUB. If you are using the Shared Inference API, you can check out this guide to follow the instructions to set up your API calls. For Dedicated Inference API users, simply click the Start Endpoint button to initiate the endpoint. Once active, HUB will give you a unique URL to use for your inference tasks.
If your project needs a model in a specific format or for offline use, Ultralytics HUB offers export options like ONNX, CoreML, or TensorFlow to support various platforms, from mobile to cloud systems. For developers looking to integrate models directly into applications, the Ultralytics HUB-SDK provides an efficient way to manage deployments through Python. By using API keys or Ultralytics credentials, you can easily control the deployment and run inferences in your code, giving you the flexibility needed for seamless integration.
Ultralytics HUB is an all-in-one platform designed to make training and deploying YOLO11 models accessible to both beginners and experts. It supports a wide range of tasks, from dataset uploads to training configuration, offering flexible deployment options like Shared and Dedicated Inference APIs. Whether you're deploying through APIs or exporting models for offline use, HUB ensures seamless integration across platforms. With options for real-time applications and scalable solutions, Ultralytics HUB can be used for a wide range of deployment needs for both beginner and advanced users.
Explore our GitHub repository and join our vibrant community to dive deeper into AI. Discover how Vision AI is advancing innovation in industries like healthcare and agriculture.
