What’s New in YOLOv11 Transforming Object Detection Once Again Part 1

Oct 18, 2024 06:35 PM - 6 months ago 249216

Introduction

At the YOLO Vision 2024 event, Ultralytics announced a caller personnel to the YOLO bid called YOLOv11. This article will supply an overview of the caller model, instructions connected really to tally conclusion utilizing YOLOv11, and the cardinal advancements and highlights of the exemplary compared to its predecessor. The YOLOv11 exemplary is designed to beryllium fast, accurate, and easy to usage for tasks specified arsenic entity detection, image segmentation, image classification, airs estimation, and real-time entity tracking. The caller state-of-the-art (SOTA) exemplary has achieved faster conclusion velocity and improved accuracy compared to the erstwhile YOLO models. Before we begin, let’s return a look astatine the benchmark results provided by Ultralytics. In the benchmark plot, the YOLOv11 exemplary has been compared to YOLOv5, v6, v7, v8, v9, and v10.

The highlighted bluish crippled is the capacity of YOLOv11 and arsenic we tin spot that it has surpassed beautiful overmuch each the yolo exemplary aliases the bid connected mean mean precision connected COCO dataset and connected conclusion velocity arsenic plotted connected the x-axis.

Tasks Supported by YOLOv11

Object Detection:- Locating objects successful an image aliases videos by drafting bounding boxes on pinch the assurance scores. Useful for applications for illustration autonomous driving, surveillance cameras, aliases postulation toll booths.
Instance Segmentation:- Identifying and segmenting objects aliases individuals successful an image. Useful for aesculapian imaging, manufacturing and more.
Pose Estimation:- Identifying cardinal points successful an image aliases video framework to show assemblage movements aliases gestures, making it useful for applications specified arsenic virtual reality, creation training, and beingness therapy.
Oriented entity discovery (OBB):- Detecting objects pinch an predisposition angle, allowing much meticulous localization of tilted aliases rotated items. This characteristic is peculiarly useful for applications specified arsenic autonomous driving, business inspection, and analyzing images from drones aliases satellites.

Model Tasks

YOLO11	Detection (COCO)
YOLO11-seg	Segmentation (COCO)
YOLO11-pose	Pose/Keypoints (COCO)
YOLO11-obb	Oriented Detection (DOTAv1)
YOLO11-cls	Classification (ImageNet)

YOLOv11 provides Detect, Segment, and Pose models pre-trained connected the COCO dataset, arsenic good arsenic Classify models pre-trained connected the ImageNet dataset. Track mode is besides disposable for each Detect, Segment, and Pose models. For much accusation astir the exemplary specifications and its different versions, please mention to the charismatic GitHub repository. We’ve included a nonstop nexus successful our resources conception for the convenience.

Prerequisites

Here are the prerequisites for moving YOLO models:

Python Environment: Install Python 3.8 aliases later.
CUDA & cuDNN: A CUDA-compatible GPU (NVIDIA) pinch CUDA and cuDNN installed for faster training and inference.
PyTorch: Install PyTorch compatible pinch your CUDA version.
YOLO Framework: Install the circumstantial YOLO type package from Ultralytics.
Dataset: Labeled dataset successful YOLO format (images and note files).
Hardware Requirements: At slightest 16 GB RAM and a GPU pinch 4+ GB VRAM for soft training and inference.

Key Feature Highlights of the caller model

YOLOv11 brings respective improvements that make it a beardown prime for machine imagination tasks. It has a amended backbone and cervix design, which helps it observe objects much accurately and grip analyzable tasks pinch ease. The exemplary is optimized for speed, offering faster processing times while still maintaining a bully equilibrium betwixt accuracy and performance. Even pinch 22% less parameters than YOLOv8m, this lightweight exemplary achieves higher accuracy, making it some businesslike and effective. YOLOv11 besides has conclusion clip 2% quicker than the YOLOv10 frankincense making it highly adaptable, moving good connected various platforms for illustration separator devices, unreality systems, and NVIDIA GPUs. Plus, it supports a wide scope of tasks, including entity detection, image classification, airs estimation, and more. YOLOv11 is designed to merge easy pinch various systems and platforms. Building connected YOLOv8’s support, it useful good successful different environments for training, testing, and deployment. Whether you usage NVIDIA GPUs, separator devices, aliases unreality platforms, YOLOv11 fits smoothly into your workflow. These features make YOLOv11 adaptable for different industries.

YOLOv11 Demo

When YOLOv11 is tally connected DigitalOcean’s GPU Droplet, the conclusion velocity reaches up to 5 to 6 sclerosis per image, making it an perfect prime for real-time applications that require accelerated and businesslike processing. We will commencement by installing the ultralytics package aliases upgrading the package.

!pip instal ultralytics --upgrade

Train the YOLOv11 exemplary for entity discovery tin beryllium done some by either Python aliases utilizing CLI commands.

from ultralytics import YOLO model = YOLO("yolo11n.pt") results = model.train(data="coco8.yaml", epochs=100, imgsz=640) yolo train model=yolo11n.pt data=coco8.yaml epochs=100 imgsz=640

We person provided the codification to usage the exemplary for detecting objects successful a video.

model = YOLO("yolo11n.pt") results = model("data/video.mp4", save=True, show=True)

Next, we will effort the exemplary to observe objects successful an image.

model = YOLO("yolo11n.pt") results = model("/folder_path/image_det.jpeg") results[0].show()

To usage the exemplary for segmentation, we request to download the YOLO11 arsenic trying the exemplary straight mightiness propulsion an error.

from ultralytics import YOLO model = YOLO('yolo11n-seg.pt') results = model("/folder_path/image_seg.jpeg") results[0].show()

Similarly, for airs estimation and classification tasks we request to download the YOLO11 exemplary and past effort the exemplary connected an image.

from ultralytics import YOLO model = YOLO('yolo11n-pose.pt') results = model("/folder_path/image_pose.jpeg") results[0].show()

from ultralytics import YOLO model = YOLO('yolo11n-cls.pt') results = model("/folder_path/image_class.jpeg") results[0].show()

Now it is advisable to usage a precocious extremity GPU to tally aliases train YOLOv11 other training aliases inferencing mightiness beryllium slow and inefficient. When it comes to moving aliases training YOLOv11, choosing GPU complete CPU tin importantly heighten capacity and efficiency. YOLOv11, pinch its enhanced characteristic extraction and improved accuracy, demands precocious computational power, particularly for training connected ample datasets. GPUs are specifically designed for parallel processing, enabling them to grip the analyzable matrix operations required for heavy learning astatine a overmuch faster complaint than CPUs. DigitalOcean GPU Droplets are optimized for AI/ML workloads, providing entree to powerful GPUs for illustration the H100, which are specially designed for precocious capacity and handling dense computing.

Concluding Thoughts

We saw immoderate cool things the exemplary tin do pinch images and videos. YOLOv11 is simply a powerful and versatile exemplary for machine imagination tasks. Its improved features and precocious velocity and accuracy make it a important upgrade complete its predecessors. In conclusion, YOLOv11 is simply a large measurement guardant successful entity discovery and machine vision. With its amended architectural design, faster speeds, and improved accuracy, it’s an fantabulous fresh for various uses—real-time discovery connected mini devices aliases much elaborate study successful the cloud. Its expertise to activity smoothly pinch existing systems intends businesses tin easy merge it into their regular operations, whether successful farming, security, aliases robotics. YOLOv11’s blend of elasticity and capacity makes it a powerful instrumentality for anyone tackling machine imagination challenges. However, this is portion 1 of the tutorial, and successful portion 2, we will study really to fine-tune and train the exemplary for entity discovery connected a civilization dataset.