In this lecture, we will explore real-world use cases that push the boundaries of modern artificial intelligence. The presentation is intended for a broad audience and showcases production scenarios where the cutting-edge AI solutions are not sufficient to solve problems. This can be due to speed, accuracy, and problem-specific constraints, or simply due to limited resources. Therefore, additional optimizations are necessary to make the solutions production-ready. During the lecture, we will uncover how classical methods and algorithms overcome these challenges through a series of practical cases, accompanied by real-world examples and demos.

We will present solutions in the fields of computer vision, image and video analysis, audio signal analysis, and solving complex optimization problems. We will discuss issues related to object tracking in videos, object detection in images, audio modification, and also touch upon classic domains such as applications in Supply Chain Management.

A special focus of the presentation is on identifying the mistakes engineers make when applying pre-trained models during system integration. Furthermore, the lecture will aim to motivate you to seek ways and opportunities to apply AI solutions to your products, regardless of your expertise in the field of AI. You will learn to recognize the need and, consequently, seek appropriate solutions.

Key Takeaways:

Stay Current with Cutting-Edge Deep Learning: Explore the latest in deep learning methods, harnessing pre-trained models while understanding their limitations.

The Power of Optimization: Prioritize optimizations during implementation to save time and resources while achieving outstanding results.

Shift Your Mindset and Embrace Classic Techniques: Embrace a mindset shift by integrating classic methods, often enhancing the quality of your solutions.

Spot Optimization Opportunities: Sharpen your ability to identify optimization opportunities within projects, unlocking peak performance potential.