More details here!
R&D Highlights
- 25 Open-Source Models: YOLO11 introduces 25 models across 5 sizes and 5 tasks, ensuring there’s an optimized model for any use case.
- Accuracy Boost: YOLO11n achieves up to a 2.2% higher mAP (37.3 → 39.5) on COCO object detection tasks compared to YOLOv8n.
- Efficiency & Speed: YOLO11 uses up to 22% fewer parameters than YOLOv8 and provides up to 2% faster inference speeds. Optimized for edge applications and resource-constrained environments.
The focus of YOLO11 is on refining architecture to improve performance while reducing computational requirements—a great fit for those who need both precision and speed.
YOLO11 Benchmarks
The improvements are consistent across all model sizes, providing a noticeable upgrade for current YOLO users.
| Model | YOLOv8 mAP (%) | YOLO11 mAP (%) | YOLOv8 Params (M) | YOLO11 Params (M) | Improvement |
|---|---|---|---|---|---|
| YOLOn | 37.3 | 39.5 | 3.2 | 2.6 | +2.2% mAP |
| YOLOs | 44.9 | 47.0 | 11.2 | 9.4 | +2.1% mAP |
| YOLOm | 50.2 | 51.5 | 25.9 | 20.1 | +1.3% mAP |
| YOLOl | 52.9 | 53.4 | 43.7 | 25.3 | +0.5% mAP |
| YOLOx | 53.9 | 54.7 | 68.2 | 56.9 | +0.8% mAP |
Versatile Task Support
YOLO11 extends the capabilities of the YOLO series to cover multiple computer vision tasks:
- Detection: Quickly detect and localize objects.
- Instance Segmentation: Get pixel-level object insights.
- Pose Estimation: Track key points for pose analysis.
- Oriented Object Detection (OBB): Detect objects with orientation angles.
- Classification: Classify images into categories.
Quick Start Example
If you’re already using the Ultralytics package, upgrading to YOLO11 is easy. Install the latest package:
pip install ultralytics>=8.3.0
Then, load a pre-trained YOLO11 model and run inference on an image:
from ultralytics import YOLO
# Load the YOLO11 model
model = YOLO("yolo11n.pt")
# Run inference on an image
results = model("path/to/image.jpg")
# Display results
results[0].show()
These few lines of code are all you need to start using YOLO11 for your real-time computer vision needs.
Access and Get Involved
YOLO11 is open-source and designed to integrate smoothly into various workflows, from edge devices to cloud platforms. You can explore the models and contribute at GitHub - ultralytics/ultralytics: NEW - YOLOv8 🚀 in PyTorch > ONNX > OpenVINO > CoreML > TFLite.
Check it out, see how it fits into your projects, and let us know your feedback!