The FlyingThings repository is a computer vision demonstration and workshop project maintained by Red Hat's North America Solutions and Sales Acceleration team. It showcases OpenShift as a platform for developing and operationalizing artificial intelligence and machine learning applications, with a specific focus on computer vision implementation using YOLOv5 for object detection. The project illustrates a complete workflow for custom model training and serving, designed for integration with other applications and services across the Red Hat ecosystem.
The repository is classified as an event-driven, cloud-native demonstration application that leverages multiple Red Hat technologies and third-party tools. It integrates Kafka for messaging, uses Tekton for pipeline orchestration, and demonstrates container orchestration capabilities on OpenShift. The architecture reflects distributed systems principles and showcases how to build microservices-based AI/ML workflows on containerized infrastructure.
The primary language of the repository is Shell, indicating that much of the automation and deployment logic is implemented through bash scripting. The project is tagged with topics including artificial intelligence, computer vision, data science, OpenShift, Tekton, YOLO, and YOLOv5, reflecting its core technical focus areas.
To run this demonstration, users need specific prerequisites including CLI tools such as bash, git, and the OpenShift command-line interface. The infrastructure requirements are substantial, necessitating OpenShift 4.12 or later with cluster-admin role access, Nvidia GPU hardware for model training and inference, and internet connectivity. Optional components include OpenShift Dev Spaces version 3.8.0 or later and AWS integration for auto-scaling capabilities.
The project provides tested deployment options through the Red Hat Demo Platform, including an AWS-based OpenShift Open Environment configuration with a control plane on m5.4xlarge instances and worker nodes on m5.2xlarge instances, as well as a single-node OpenShift option on m5.4xlarge. An MLOps-focused variant is also available for organizations interested in data science and edge computing practices.
The repository includes comprehensive workshop documentation in docs/WORKSHOP.md, accompanied by a video series that explains the workshop content and provides walkthroughs of the setup and use cases. Users can explore the demonstration either through OpenShift Dev Spaces for AI and ML, which is available as a separate companion project, or on their local machines. The quickstart section directs users to the workshop instructions for detailed implementation guidance.
The project serves as both an educational resource and a practical reference implementation for organizations looking to deploy computer vision applications on OpenShift. It demonstrates how to integrate custom-trained machine learning models with containerized services, manage complex AI/ML workflows through Tekton pipelines, and leverage Kafka for event-driven communication between components. The inclusion of multiple deployment options and the emphasis on the Red Hat ecosystem make it particularly valuable for enterprises already invested in Red Hat technologies or considering OpenShift for their AI/ML infrastructure needs.
