Welcome to the Gwel Wiki!

The gwel Python module provides a framework for handling large image datasets and using neural networks for computer vision tasks for scientific research. The module supports object instance detection and semantic segmentation. The flowchart below outlines the workflows that can be achieved using this module.

This wiki compiles a series of tutorials designed to guide you through the features and functionality of the module. For the best learning experience, start with the first tutorial and proceed sequentially.

The name Gwel is derived from Gweles (Cornish: to see).

Maintained and created by Jack Rich ([email protected]), Department of Crop Science, School of Agriculture, Policy, and Development; University of Reading as part of my PhD research.

📘 Tutorials Overview

1. Installation and Command Line Interface

This tutorial provides a brief introduction to the package and includes the installation guide found in the project’s README.md.

2. Handling Image Datasets using the ImageDataset Class

Learn to handle collections of images with the ImageDataset class. Key operations include resizing images, taking samples, orienting, verifying integrity, and viewing images.

3. Handling Image Datasets with Factors

Add factors (additional variables associated with images, e.g., time or treatment) to the ImageDataset class. Efficiently manage multiple factors in large image collections.

4. Object and Instance Detection with Pre-trained Models

Introduces the Detector class to run pre-trained object detection models and store results in an ImageDataset. Covers visualization, exporting results in COCO JSON format, and cropping images to create individual object datasets.

5. Semantic Segmentation with Pre-trained Models

Introduces the Segmenter class to run pre-trained image segmentation models. Covers visualization and exporting segmentation results in COCO JSON format.

6. Image Annotation and Model Training

Learn to annotate images using CVAT, then train object detection or segmentation models. Includes tips for leveraging GPU clusters for faster training.

⚡ Tips for Using This Wiki

• Follow tutorials in order for the best learning experience.
• Experiment with sample datasets before using your own.
• Use the links above to jump directly to a tutorial of interest.

Acknowledgments

This research is funded by the Biotechnology and Biological Sciences Research Council (BBSRC), part of UK Research and Innovation (UKRI), through the FoodBioSystems Doctoral Training Partnership (DTP) as part of my PhD project at the University of Reading.