Artificial Intelligence (AI) has been revolutionary, but its rapid growth has come with significant environmental costs. Training and deploying large-scale AI models require substantial computational resources, resulting in increased energy consumption and greenhouse gas emissions. This tutorial addresses these challenges by introducing energy-efficient deep learning practices and tools to measure, analyze, and mitigate the carbon footprint of AI systems.

The goals of this tutorial are:

1. Educate participants on the environmental impact of AI and the importance of sustainable practices.
2. Provide hands-on experience with tools for monitoring and reducing carbon emissions during AI model training.
3. Explore energy-efficient implementations of object detection models such as YOLO-v5 and R-CNN.

By the end of the tutorial the attendees are able to:

• Understand Environmental Impact: Gain insights about the importance of sustainable AI techniques as well as the energy usage and carbon emissions of AI models.
• Utilize Carbon Monitoring Tools: Learn how to track and lower the carbon impact of AI tasks by utilizing tools such as CodeCarbon, CarbonTracker, and Eco2AI.
• Compare Model Efficiency: Gain the ability to evaluate the energy efficiency of various AI models and training environments.
• Implement Object Detection Models: Acquire hands-on experience in training and evaluating object detection models, specifically YOLO-v5 and R-CNN, while applying energy-efficient practices.
• Promote Sustainable AI Practices: Understand how to incorporate energy-efficient techniques into AI processes to support ecologically conscious and responsible innovation.

Typical Energy Measurement Output

GitHub Repository

• Climate Change
• AI
• Machine Learning