TechMeet.AI - Sotogrande, 14 June 2024

As artificial intelligence and machine learning technologies transition from the domain of cutting-edge research to more accessible tools, they have increasingly become commodities that empower developers, businesses, and researchers. This talk will explore the commoditization of AI, focusing on the ease of accessing, fine-tuning, and deploying pre-trained models. We will delve into how platforms like Hugging Face have revolutionized the field by providing open access to a multitude of models that can be easily adapted to fit specific needs. In this session, we will provide a practical demonstration of utilizing Hugging Face transformer models and Python libraries to train a Text-to-SQL model. Attendees will learn how to select a suitable pre-trained model, fine-tune it on a specific dataset, and effectively deploy it into a functional application.

In the era of swiftly advancing machine learning models, computer vision emerges as an area of particular intrest. It is a vast field of potential applications, ranging from tumor detection and facial recognition to geospatial analysis and hazard identification. The forthcoming discourse shall center upon a specialized subdomain within computer vision: real-time, open-vocabulary object detection. This domain represents a convergence of computer vision and natural language processing methodologies, yielding vision-language multimodal models. The initial segment of this presentation will be focused on the mathematical foundation of these models and through visualization and model structural analysis, we shall garner insight into the operational mechanisms of each constituent element and its rationale. The presentation will be facilitated through a application tailored specifically for this purpose. Consequently, participants will be equipped with both theoretical understanding of vision language models and practical know-how of computer vision models implementation in Python programming language.

Leave a Comment