Generative AI is rapidly reshaping how interactive digital content is created, making it an essential skill for professionals in information engineering working with graphics, multimedia, and simulations. From realistic digital humans to procedurally generated virtual worlds, generative models are becoming the creative engine behind next-generation media pipelines. The course goal is to provide the student with both practical and theoretical content on generative models and their application in the context of computer graphics, multimedia, and interactive simulations. The course introduces modern deep generative models such as autoregressive networks, GANs, diffusion models, and reinforcement learning applied to generative tasks. The course will provide the student with the ability to understand and implement modern generative models, and to apply them in real-world scenarios such as digital human generation, procedural content creation, and interactive simulations.

At the end of the course, the students will achieve basic knowledge pertaining: - principles of autoregressive models, VAEs, GANs, normalizing flows and diffusion models; - generative reinforcement learning techniques; - state-of-the-art tools and frameworks for implementing generative architectures; - applications of generative AI in multimedia, animation, and computer graphics; - methods for combining learning-based generation with simulation and interactivity. The student will be able to: - understand the differences between different generative models and select the optimal models based on application requirements; - train or finetune a deep generative model on a custom dataset for conditional and unconditional synthesis of images and graphics content; - integrate generative models into creative and interactive pipelines; - use reinforcement learning to drive generative processes and simulations.

Requirements: - Fundamentals of computer programming. - Mathematical analysis I. Linear algebra. - Basic knowledge of machine learning and neural networks. - Basic understanding of computer graphics and animation. Fundamentals of Unity and Blender are suggested.

The tentative course program will be articulated in in-class theoretical lessons and lab exercises, as reported below. - Autoregressive models (in-class lessons 3h + lab exercises 3h) - Generative Adversarial Networks (GANs) (in-class lessons 3h + lab exercises 3h) - Variational Autoencoders (VAEs), Normalizing flows and latent diffusion models (in-class lessons 12h + lab exercises 3h) - Reinforcement Learning for generative AI (in-class lessons 9h + lab exercises 4h) * Use of RL for generative sequences, behaviors, and animation * ML Agents and interactive environment simulation - Virtual humans and generative character design (in-class lessons 4h + lab exercises 3h) * Generation of faces, bodies, and expressions using VAEs and GANs * Custom digital humans for graphics and simulation - Generative AI for procedural environment creation (in-class lessons 3h + lab exercises 4h) * Procedural generation of 3D layouts, terrains, and textures * Use of generative models for scenario design * Scene representations with neural fields and Gaussian Splatting - Final project (6h)

The course will encompass 40 hours of in-class lessons and 20 hours of lab exercises. Exercises will be aimed at introducing the software stack used for the implementation of neural networks and the practical aspects associated with their training. During lab activities, students will be provided with problems to solve individually or in a pair, which will be followed by a discussion of the main approaches adopted to reach the solution. These activities will be preparatory to the development of an individual or group project, which will concur to the determination of the final grade. Seminars could be organized to tackle advanced topics.

Slides and supplementary material, including lab exercises and solutions, are provided by the teachers.

Slides; Esercitazioni di laboratorio; Esercitazioni di laboratorio risolte; Materiale multimediale ;

Modalità di esame: Prova orale obbligatoria; Elaborato progettuale in gruppo;

Exam: Compulsory oral exam; Group project;

... The assessment will consist of a group project (25/30) and an oral exam (5/30), as detailed below. Group project. This part focuses on the design and implementation of a solution to a practical problem in the domain of generative AI for computer graphics and multimedia. Example topics include the generation of 3D environments, digital humans, expressive avatars, textures, or interactive behaviors for agents and NPCs. Students will work in groups (2-3 students) to select or define a problem and then reproduce and adapt a deep learning method to address it effectively. Each group will submit an 8-page PDF report in IEEE paper format, due 10 days before the oral exam. This part of the exam is intended to evaluate the students' ability to implement and experimentally validate a deep learning model, as well as to report and critically discuss the obtained results. The evaluation includes an assessment of the submitted report. Extra points (for honors, "lode") will be awarded to projects that propose original extensions, that tackle particularly challenging problem settings, or that include well-justified and carefully executed experiments. Oral Presentation and Theory questions. This part of the exam aims at verifying the equal contribution of all the authors to the project work. The students are asked to present the work done. The content and timing of the presentation will be evaluated as well as the answers to individual questions. Students are also asked one question about more general theory topics presented in class. Linee guida Up to +4 points can be awarded for the submission of labs/homeworks or for extra activities organized in the context of the Master's Degree, pertaining to the course content.

Gli studenti e le studentesse con disabilità o con Disturbi Specifici di Apprendimento (DSA), oltre alla segnalazione tramite procedura informatizzata, sono invitati a comunicare anche direttamente al/la docente titolare dell'insegnamento, con un preavviso non inferiore ad una settimana dall'avvio della sessione d'esame, gli strumenti compensativi concordati con l'Unità Special Needs, al fine di permettere al/la docente la declinazione più idonea in riferimento alla specifica tipologia di esame.