Segmentation Tutorial

[paulguerrie]

Added a tutorial.ipynb to segment/ to demonstrate training, validation, and prediction for YOLOv5 Segmentation.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

New Jupyter notebook tutorial for YOLOv5 segmentation capabilities.

📊 Key Changes

• Addition of a comprehensive tutorial notebook (tutorial.ipynb) for YOLOv5 segmentation.
• Tutorial covers setup, prediction examples, validation on COCO dataset, training, and visualization.

🎯 Purpose & Impact

• Ease of Use: Allows users to follow step-by-step instructions to use YOLOv5 for segmentation tasks.
• Educational Value: Helps new users understand YOLOv5's segmentation process, from installation to training and inference.
• Performance Verification: Users can validate model accuracy on standard segmentation datasets, ensuring reliability and robustness.
• Customizability: The tutorial demonstrates training on custom datasets with Roboflow, fostering adaptation to diverse usage scenarios.
• Reproducibility: Integrated logging and third-party platforms (Comet, ClearML) support experiment tracking and model training reproducibility.

[glenn-jocher]

@paulguerrie awesome, looks much better :)

[glenn-jocher]

Uploading result for use in notebook

[Sarveshsh8]

How to run a inference just like object detection?
model = torch.hub.load()
is there any method to run a inference like this for segmentation?

[glenn-jocher]

@Sarveshsh8 yes, for segmentation you can run inference by loading a model and passing inputs through it. Here's an example using the YOLOv5 model:

import torch
from PIL import Image

# Load the model
model = torch.hub.load('ultralytics/yolov5', 'custom', path_or_model='best.pt')

# Load an image
img = Image.open('path/to/image.jpg')

# Perform inference
results = model(img)

# Display results
results.show()

For more detailed usage, check out the official Ultralytics YOLOv5 documentation.