Welcome! This is the home page of the "Advanced Machine Learning and Computer Vision" course, taking place in the winter semester of 2026/2027.
The course is part of the Master's Degree in Data Science -- Sapienza University of Rome -- organized jointly by the departments of Computer Science (DI), Information and Automation Engineering (DIAG) and Information, Electronics and Telecommunication Engineering (DIET), and Statistics (DSS).
Important: subscribe to the course mailing list to receive slides, course material, assignments and news.
Click on this link.
Note: join the mailing list with your institutional email.
If you did not receive your institutional address yet due to specific situations, please request access to the group but also send an email detailing i. the circumstance; ii. the proof of your acceptance; and iii. the proof of your identity. Requests from non-institutional addresses without those cannot be accepted.
Classes starts on September, 25th (in agreement with the Data Science and Computer Science class schedule).
The course takes place @:
Room B2 (Via Ariosto, 25) on Mondays 12:30-15:00
Room B2 (Via Ariosto, 25) on Wednesdays 16:00-18:00
Please also refer to the Data Science and Computer Science first-semester lecture times for the class schedule.
Classes are in presence.
Prof. Fabio Galasso, web-page, email: galasso remove_this @di DOT uniroma1 DOT it
The course presents advanced concepts of machine learning and their application to computer vision and embodied AI, via deep neural networks and foundation models. It includes theory and practical coding in PyTorch, an assignment, and a final hands-on project.
In a first part, after a compact recap of the deep learning building blocks (convolutional networks, attention and Transformers), I will address visual perception: semantic segmentation and object detection, from region-based detectors to end-to-end Transformer detectors such as DETR, and human-centric understanding, including pose estimation and human motion forecasting. I will then discuss the modelling of sequences and structured data, with Graph Convolutional Networks for skeletons and interacting agents, and State-Space Models for long-term temporal modelling.
In a second part, I will discuss representation learning, generalization and the effective use of labelled and unlabelled data. Further to transfer learning, I will present self-supervised learning, from pretext tasks to contrastive and positive-only approaches, and multi-modal learning that aligns vision with language and other sensor modalities (e.g. CLIP). I will introduce hyperbolic deep learning for hierarchical representations, showcasing its use for estimating uncertainty, for active learning, where the model requests the intervention of humans, and for domain adaptation (e.g. applying daytime detectors to night vision).
In a third part, I will discuss generative AI and foundation models. I will cover VAEs, GANs, diffusion and flow matching models, with applications such as motion synthesis and anomaly detection. I will then present Vision Language Models (VLMs): how visual encoders are connected to large language models, how they are instruction-tuned, and how they support visual reasoning.
Finally, I will move from perception to action. I will introduce world models, which learn to predict how the environment evolves, world action models, which jointly anticipate future states and the actions to take, and Vision Language Action models (VLAs), which map visual observations and language instructions to robot actions. I will ground these in robotics, including manipulation, social navigation among people and human-robot interaction. Finally, I will address safe AI through machine unlearning (removing concepts or data from trained models), model editing, and steering (guiding model behaviour at inference time, e.g. anticipating and preventing an agent's failure before it acts). These topics draw on recent research of our group, published at top venues such as ICLR, ICML, ECCV, AAAI and IROS.
The students' expertise will be assessed at the first two lectures, to set the course at the most advanced possible starting point.
The following are pre-requisites which attendants should possess at the course start, or be ready to remedy with own study:
Proficiency in Python, some experience with Pytorch
Calculus, Linear Algebra, Probability
Taking derivatives, understanding matrix vector operations and notation
Machine Learning
Regression, binary and multi-class classification
Cost functions, derivatives and optimization with gradient descent
Basics of Deep Learning
Basics of fully-connected and convolutional networks
Basics of backprop
Suggestions and material for own study to remedy pre-requites will be discussed upon request.Â