Ultralytics: Open-Source Computer Vision & Deep Learning Framework 149

Ultralytics is a useful open-source computer vision and deep learning framework. It is primarily used for developing and running the YOLO series of models, such as YOLOv5, YOLOv8, and others. This tool is commonly used in vision tasks that require fast response, such as object detection, image segmentation, and action recognition.

Simply put, it is a toolkit that makes it easy to go from training a model to using it in practice.

Ultralytics Core Features

Supports Multiple Tasks

It can handle a wide range of computer vision tasks, including:

• Object detection: mark the position of objects.

• Image segmentation: precisely cut out the outline of objects.

• Pose estimation: identify the positions of a person's joints.

• Classification: determine what the entire image is.

• Oriented detection: find tilted objects, such as angled cars in aerial imagery.

Full Development Pipeline

• It manages the entire process, from data preparation, model training, to final export and use.

Fast and Easy to Learn

• Simple to use: offers straightforward code commands, easy for beginners to learn.

• Runs fast: supports various acceleration techniques for very fast computation.

• Multiple models: offers both large and small models, depending on whether you need speed or accuracy.

Easy to Deploy Anywhere

• Models can be converted into various formats and run on computers, mobile phones, or dedicated hardware devices.

Ultralytics Use Cases

• Factory quality inspection: check product defects, count parts, read codes — already used on more than 200 production lines.

• Urban security: face recognition, vehicle monitoring, fire detection, and traffic flow counting.

• Robots and vehicles: obstacle avoidance, road perception, and lane recognition.

• Healthcare and research: cell segmentation, X-ray image analysis, and experimental testing.

• Small devices: drones, surveillance cameras, and handheld devices — small models can also run on regular processors.

How to Use Ultralytics

Ultralytics offers flexible ways to use it, whether through Python code or the command line.

Python API Usage Example

from ultralytics import YOLO

# 1. Load a pre-trained model (e.g., YOLOv8 nano)
model = YOLO("yolov8n.pt")

# 2. Train the model on a custom dataset
model.train(data="coco.yaml", epochs=100, imgsz=640)

# 3. Perform object detection on an image
results = model("image.jpg")

# 4. Save the results
results[0].save(filename="result.jpg")

Command-Line Interface (CLI) Examples

# Predict on an image
yolo predict model=yolov8n.pt source='path/to/image.jpg'

# Start training the model
yolo detect train data=coco.yaml model=yolov8n.pt epochs=100

Common Questions & Troubleshooting