A few days ago, Ultralytics surprised us again with a major announcement. They introduced their new model, YOLOv11, at a spectacular event that showcased impressive advancements in technologies such as computer vision and machine learning. This new version not only marks a milestone but redefines what can be achieved in the field of computer vision.
YOLOv11 is the result of collaboration between AI experts, innovators, and developers, designed to take computer vision capabilities to new heights. With a range of technical innovations, this model is faster, more accurate, and more versatile, positioning itself as an essential tool for both developers and researchers.
Its architecture makes it the ideal solution for a variety of computer vision tasks, from real-time object detection to classification. Key improvements include enhanced feature extraction, increased accuracy with fewer parameters, and faster processing speeds. In this article, we will explore how YOLOv11 significantly boosts real-time performance and opens up new possibilities for advanced applications.
Why YOLO11 instead of YOLOv11?
I don’t know, but this is the answer I received from Muhammad Rizwan Munawar at Ultralytics, so we’ll have to find out!
Main features of YOLO11: Speed and accuracy in computer vision
The new gem in the Ultralytics family improves on everything we’ve seen in previous versions, offering unparalleled precision and speed. This model stands out for its architectural enhancements and advancements in training methods, making it a powerful option for various computer vision applications.
How YOLO11 Performs with the COCO Dataset
The Common Objects in Context (COCO) dataset is the standard benchmark for evaluating object detection models, using the widely recognised mean Average Precision (mAP) metric to measure both precision and recall.
In this context, YOLO11 represents a major leap in object detection technology. Thanks to its innovative architecture and enhanced training optimisations, it outperforms its predecessors and sets a new state-of-the-art standard. The YOLO11 nano version (YOLO11n) achieves a mAP of 39.5, while the Extra Large version (YOLO11x) reaches 54.7.
Key Features of YOLO11
- Advanced Feature Extraction: YOLO11 boasts a revamped architecture, enabling it to capture fine image details more efficiently. This enhancement allows for higher accuracy in object detection, even in complex scenarios.
- Optimised for Speed and Efficiency: With improved architectural design and optimised training channels, YOLO11 processes images faster than its predecessors while maintaining high accuracy.
- Higher Precision with Fewer Resources: YOLO11 achieves superior mean Average Precision (mAP) using fewer parameters than YOLOv8, making it both more accurate and resource-efficient.
- Versatility Across Different Environments: Whether deployed on edge devices, cloud platforms, or NVIDIA GPU systems, YOLO11 seamlessly adapts to various settings, making it suitable for small-scale projects and large-scale solutions alike.
- Broad Range of Supported Tasks: From real-time object detection to complex tasks like instance segmentation, image classification, pose estimation, and oriented bounding boxes (OBB), YOLO11 is a powerful tool for developers and researchers.
Maximise the Potential of YOLO11 and Master Its Key Functionalities
Ultralytics YOLO11 is more than just another object detection model—it’s a complete ecosystem covering the entire machine learning lifecycle. From data ingestion to deployment and real-time monitoring, YOLO11 offers a range of modes designed to provide flexibility and efficiency at every stage of your project.
Key modes of YOLO11
Training mode: Fine-tunes the model on custom datasets, optimising parameters for accurate predictions.
Validation mode: Assesses post-training accuracy and generalisation, ideal for fine-tuning.
Prediction mode: Uses trained models for real-world inference on images or videos.
Export mode: Converts models for production environments, ensuring compatibility across platforms and devices.
Tracking mode: Enables real-time object tracking, useful for surveillance and autonomous vehicles.
Benchmark mode: Evaluates speed and accuracy across different export formats to identify the most efficient option.
Train, validate, predict, and export with ease
- Training: Fine-tune your model using specific data and parameters to optimise object detection accuracy.
- Validation: Assess your model’s performance post-training to ensure strong generalisation and precision.
- Prediction: Perform real-time predictions or analyse new image and video datasets with a trained model.
- Export: Prepare your model for deployment in multiple formats, ensuring compatibility across different systems and devices.
Prediction on videos, images, or cameras with YOLO11
At the time of writing, and in my opinion, Ultralytics is the most versatile and useful computer vision framework. It provides an accessible and efficient way to perform inference or train neural networks with just a few lines of code, making it an invaluable tool for professionals and researchers in the field.
Below, we’ll show you how to harness YOLO11’s potential to run inference on different data sources in just a few simple steps:
Python 🐍
```python
from ultralytics import YOLO
# Load a model
model = YOLOv10('yolo11x.pt') # load an official model
# Predict with the model
model.predict(0) # predict on your webcam
model.predict(“video.mp4”) # predict on video
model.predict(“rtsp://user:password@ipcamera:portcamera”) # predict on rtsp camera
model.predict(“image.jpg”) # predict on image
```CLI 🤓
“`bash
yolo predict detect model=yolo11x.pt source=video.mp4 show=True save=True
“`
YOLO11 Results and comparison with other YOLO models
It’s incredible how fast the YOLO model has advanced since the launch of YOLOv5 by Ultralytics. The differences between the last four versions (YOLOv8 to YOLO11) are almost imperceptible, highlighting the remarkable development and continuous evolution in object detection. These improvements have benefited thousands of users, making advanced computer vision solutions more accessible and widely available to anyone looking to implement them.
Supported tasks and modes in YOLO11
Below is a table highlighting the different versions of the model and their compatibility with inference, validation, and export modes. This showcases YOLO11’s ability to handle a wide range of computer vision tasks, from basic object detection to advanced segmentation and classification.
Capabilities enabled by YOLO11
This model is designed to tackle a wide range of tasks, offering advanced support for various applications. Key features include:
- Object detection: Identifies and highlights objects in images or videos, essential for security, autonomous vehicles, and retail analytics.
- Instance segmentation: Isolates and distinguishes objects at the pixel level, useful in medical imaging and quality control.
- Image classification: Categorises images, aiding e-commerce and wildlife monitoring.
- Pose estimation: Tracks key points for sports analysis and healthcare.
- Oriented Bounding Boxes (OBB): Detects objects at specific angles, ideal for robotics and aerial imagery.
- Object tracking: Monitors movement across frames in real-time, crucial for security and traffic management.
Reflection
YOLO models continue to impress, and while YOLOv10 made its mark, YOLO11 has established itself as one of the most advanced in computer vision. Currently, its development focuses on enhancing both accuracy and processing speed, making it even more efficient for real-time applications. With key advancements in object detection in challenging conditions, YOLO11 is opening up new opportunities in areas like security, industrial automation, and autonomous driving, where speed and precision are crucial.
If you’re interested in learning how YOLO11 and computer vision technology are transforming industries, speak to our team to explore how these innovations can enhance efficiency, speed, and security in your sector.
