Understand how Ultralytics YOLO11 can enhance object detection using oriented bounding boxes (OBB) and what applications this computer vision task is ideal for.
Ultralytics’ annual hybrid event, YOLO Vision 2024 (YV24), focused on discussing the latest breakthroughs in AI and computer vision. This was the perfect occasion to introduce our newest model, Ultralytics YOLO11. The model supports the same computer vision tasks as Ultralytics YOLOv8, making the shift to the new model effortless for users.
Suppose you were using YOLOv8 for oriented bounding boxes (OBB) object detection to detect objects from various angles. You can now switch to YOLO11 with a few minor changes to your code and benefit from YOLO11’s improvements ranging from increased accuracy and efficiency to processing speed. In case you have yet to use models like YOLO11, OBB detection is a great example of how YOLO11 can be applied across a range of industries, offering practical solutions that make a real impact.
In this article, we’ll look at what OBB object detection is, where it can be applied, and how to use YOLO11 to detect OBB. We'll also walk through how YOLO11's new features can improve these processes and how to run inferences and train custom models to make the most of its OBB detection capabilities.
OBB object detection takes traditional object detection a step further by detecting objects at different angles. Unlike regular bounding boxes that stay aligned with the image’s axes, OBBs rotate to fit the object’s orientation. OBB object detection can be used to analyze aerial or satellite images where objects aren’t always straight. In industries like urban planning, energy, and transportation, the ability to accurately detect angled objects like buildings, vehicles, or infrastructure can form the foundation of computer vision applications with tangible benefits.
YOLO11 supports OBB detection and was trained on the DOTA v1.0 dataset to detect objects like planes, ships, and storage tanks from different perspectives. YOLO11 comes in several model variations to suit different needs, including YOLO11n-obb (Nano), YOLO11s-obb (Small), YOLO11m-obb (Medium), YOLO11l-obb (Large), and YOLO11x-obb (Extra Large). Each model offers a different size, with varying levels of speed, accuracy, and computational power. Users can choose the model that offers the right balance of speed and accuracy for their application.
YOLO11’s object detection capabilities, especially its support for oriented bounding boxes, bring greater precision to various industries. Next, we'll take a look at a few examples of how YOLO11 and OBB detection can be used in real-world situations to make processes more efficient, accurate, and easier to manage in different fields.
If you've ever admired the design and layout of a city, it's thanks to the detailed work of urban planning and infrastructure monitoring. One of the many aspects of infrastructure monitoring is identifying and managing important structures like storage tanks, pipelines, and industrial sites. YOLO11 can help city planners analyze aerial images to detect these critical components quickly and accurately.
Oriented bounding box object detection is particularly helpful here because it enables the detection of objects viewed from various angles (often the case with aerial imagery). Precision is vital here to keep track of industrial zones, manage environmental impacts, and ensure infrastructure is properly maintained. OBB makes the detection process more reliable, helping planners make informed decisions about the city’s growth, safety, and sustainability. Using YOLO11, can help planners monitor and manage the infrastructure that keeps cities running smoothly.
As renewable energy and innovations like solar farms become more popular, regular inspections are becoming more important. Solar panels need to be checked to make sure they're working efficiently. Over time, things like cracks, dirt buildup, or misalignment can lower their performance. Routine inspections help catch these problems early, so maintenance can be done to keep them running smoothly.
For instance, solar panels can be inspected for damages using drones that are integrated with edge AI and YOLO11. Analyzing images on the edge brings more precision and efficiency to the inspection process. Due to the drone's movement and perspective, the surveillance footage may often capture solar panels from various angles. In these cases, YOLO11’s OBB detection can help drones accurately identify solar panels.
Ports and harbors handle hundreds of ships each week, and managing such a large fleet can be challenging. An added element of difficulty is involved when analyzing ships in aerial images; the ships often appear at different angles. This is where YOLO11’s support for OBB detection comes in handy.
OBB detection makes it possible for the model to detect ships at various angles more accurately than standard rectangular boxes. By using YOLO11 with OBB, shipping companies can more easily identify their fleet's location and condition, keeping track of important details like fleet movements and supply chain logistics. Such vision-enabled solutions help optimize routes, reduce delays, and improve overall fleet management across shipping routes.
If you're an AI developer looking to use YOLO11 for OBB detection, there are two easy options to get started. If you're comfortable working with code, the Ultralytics Python package is a great choice. If you prefer a user-friendly, no-code solution with cloud training capabilities, Ultralytics HUB is an in-house platform designed just for that. For more details you can take a look at our guide on training and deploying Ultralytics YOLO11 using Ultralytics HUB.
Now that we’ve seen examples of where YOLO11’s OBB support can be applied, let’s explore the Ultralytics Python package and see how you can run inferences and train custom models using it.
First, to use YOLO11 with Python, you'll need to install the Ultralytics package. Depending on your preferences, you can choose to install it using pip, conda, or Docker. For step-by-step instructions, you can refer to our Ultralytics Installation Guide. If you face any challenges during installation, our Common Issues Guide offers helpful troubleshooting tips.
Once you’ve installed the Ultralytics package, working with YOLO11 is incredibly straightforward. Running an inference refers to the process of using a trained model to make predictions on new images - like detecting objects with OBB in real time. It is different from model training, which is when you teach the model to recognize new objects or improve its performance on specific tasks. Inferencing is used when you want to apply the model to unseen data.
The example below walks you through how to load a model and use it to predict oriented bounding boxes on an image. For more detailed examples and advanced usage tips, make sure to check out the official Ultralytics documentation for best practices and further instructions.
Training a YOLO11 model means you can fine-tune its performance on specific datasets and tasks, such as oriented bounding box object detection. While pre-trained models like YOLO11 can be used for general object detection, training a custom model is essential when you need the model to detect unique objects or optimize performance on a specific dataset.
In the code snippet below, we cover the steps for training a YOLO11 model for OBB detection.
First, the model is initialized using pre-trained YOLO11 OBB-specific weights (yolo11n-obb.pt). Then, a training function is used to train the model on a custom dataset, with parameters such as the dataset configuration file, the number of training cycles, the training image size, and the hardware to run the training on (e.g., CPU or GPU). After training, the model's performance is validated to check metrics such as accuracy and loss.
Using the trained model, you can run inferences on new images to detect objects with OBB and visualize them. Also, the trained model can be converted to formats like ONNX for deployment using the export feature.
Ultralytics YOLO11 takes object detection to the next level with its support for oriented bounding boxes. By being able to detect objects at different angles, YOLO11 can be used for various applications across different industries. For example, it is a perfect fit for industries like urban planning, energy, and shipping, where precision is crucial for tasks such as solar panel inspections or fleet monitoring. With faster performance and improved accuracy, YOLO11 can help AI developers solve real-world challenges.
As AI becomes more widely adopted and integrated into our daily lives, models like YOLO11 will shape the future of AI solutions.
To learn about Vision AI’s potential, join our growing community and explore our GitHub repository. We are driving innovation by reinventing sectors like healthcare to self-driving cars.
Bắt đầu hành trình của bạn với tương lai của machine learning
