Artificial Intelligence and Machine Learning are popular topics at the moment, but how much of it is hype, and can AI deliver tangible business benefits? We can use the pre-trained models in Azure Cognitive Services to identify well-known objects and landmarks, but that's a narrow field that's the result of huge research projects and ingesting massive amounts of data. How can we apply that to practical business problems, like visually detecting faults, spotting our company logo, or our competitor's products on shelves?
To understand the answer to this question you need to understand the principle of Transfer Learning. We'll look at the theory, and then see it working, and see some of the tools in Azure Cognitive Services and other toolkits that allow us to use this technique without needing to be an expert in AI. In fact you don't even need to be able to code.
When you first see Transfer Learning in action it seems like a miracle - how could it possibly work? I'll try to show how it's really just common sense. It also gives us some insight into how human learning and intelligence works, and perhaps a glimpse into the future. In this one talk we're going to cover some of the theory behind neural networks, see a practical demonstration of its application to real problems, and in the time remaining, see the future of humanity!
You will learn:
How to visualize neural networks using libraries like TensorFlow and PyTorch
How pre-trained neural networks can be adapted to new applications using Transfer Learning
How to build custom AI models using Azure Cognitive Services