Machine supercharges your GitHub Workflows with seamless GPU acceleration. Say goodbye to the tedious overhead of managing GPU runners and hello to streamlined efficiency. With Machine, developers and organizations can effortlessly scale their AI and machine learning projects, shifting focus from infrastructure headaches to innovation and speed.

GPU-Accelerated Object Detection

This repository provides a complete, automated example of performing GPU-accelerated object detection using the DETR model (facebook/detr-resnet-50) and GitHub Actions powered by Machine.dev. It leverages Hugging Face Transformers to detect and annotate objects from images in the COCO2017 dataset.

---

✨ Key Features

• ⚡ GPU Acceleration: Efficiently perform object detection using GPUs via Machine.dev.
• 📥 Seamless Data Integration: Automatically fetch metadata and images from the phiyodr/coco2017 dataset.
• 🔍 Automated Detection Pipeline: Detect, annotate, and export results without manual intervention.
• 📂 Results in CSV: Generate a structured CSV artifact with detection details for easy review.
• 🚀 Easy Deployment: Use this repository as a GitHub template to quickly start your own GPU-accelerated workflows.

---

📁 Repository Structure

├── .github/workflows/
│   └── batch-object-detection.yml  # Workflow definition
└── object_detection.py             # Main detection script

---

▶️ Getting Started

1. Use This Repository as a Template

Click the Use this template button at the top of this page to create your own copy.

2. Set Up GPU Runners

Ensure your repository uses Machine GPU-powered runners. No additional configuration is required if you're already using Machine.dev.

3. Run the Workflow

• Trigger via commit or manually using GitHub Actions (workflow_dispatch).
• The workflow downloads and processes the first 1000 images from COCO2017.

You can run the benchmarking workflow manually via GitHub Actions using the workflow_dispatch trigger. This allows you to input parameters such as the tenancy you wish to use:

on:
  workflow_dispatch:
    inputs:
      tenancy:
        type: choice
        required: false
        description: 'The tenancy of the machine'
        default: 'spot'
        options:
          - 'spot'
          - 'on_demand'

These inputs are configurable directly in the GitHub Actions UI when manually triggering the workflow.

The workflow itself is fairly routine:

jobs:
  detect_objects:
    name: Detect Objects
    runs-on:
      - machine
      - gpu=T4
      - cpu=4
      - ram=16
      - architecture=x64
      - tenancy=${{ inputs.tenancy }}
    steps:

      - uses: actions/checkout@v4

      - name: Set up Python 3.10
        uses: actions/setup-python@v4
        with:
          python-version: '3.10'

      - name: Install Dependencies
        run: |
          pip install -r requirements.txt

      - name: Run Object Detection
        run: python3 object_detection.py

      - name: Upload Detection Results CSV
        uses: actions/upload-artifact@v4
        with:
          name: detection-results-csv
          path: detection_output/detections.csv

We have configured the detect_objects job to run on a Machine GPU runner with the desired specs.The workflow then installs dependencies, runs the object detection script, and uploads the resulting CSV as an artifact.

4. Review Outputs

• Annotated images and detections.csv will be saved in the detection_output/ directory.
• The CSV file is uploaded as a GitHub Actions artifact for convenient access.

---

🔑 Prerequisites

• GitHub account
• Access to Machine GPU-powered runners

No local installation necessary—all processes run directly within GitHub Actions.

---

📄 License

This repository is available under the MIT License.

---

📌 Notes

• This repository is currently open for use as a template. While public forks are encouraged, we are not accepting Pull Requests at this time.

For questions or concerns, please open an issue.