Machine Perception

Abstract

Recent developments in neural networks have drastically advanced the performance of machine perception systems in a variety of areas including computer vision, robotics, and human shape modeling.This course is a deep dive into deep learning algorithms and architectures with applications to a variety of perceptual and generative tasks.

Objective

Students will learn about fundamental aspects of modern deep learning approaches for perception and generation. Students will learn to implement, train and debug their own neural networks and gain a detailed understanding of cutting-edge research in learning-based computer vision, robotics, and shape modeling. The optional final project assignment will involve training a complex neural network architecture and applying it to a real-world dataset. The core competency acquired through this course is a solid foundation in deep-learning algorithms to process and interpret human-centric signals. In particular, students should be able to develop systems that deal with the problem of recognizing people in images, detecting and describing body parts, inferring their spatial configuration, performing action/gesture recognition from still images or image sequences, also considering multi-modal data, among others.

Content

The courses focuses on teaching how to set up the problem of machine perception and the associated learning algorithms, neural network architectures, and advanced deep learning concepts. The course covers the following main areas: I) Foundations of Deep Learning: Multilayer perceptrons, backpropagation, time-series modeling, convolutional neural networks. II) Advanced topics: latent variable models, generative adversarial networks, auto-regressive models, invertible neural networks, normalizing flows, diffusion models, neural implicit surface representations, neural radiance fields. III) Applications in machine perception and human-centric computer vision: general understanding of human activities, 3D reconstruction of human performance using different input modalities (monocular or multi-view images, body-worn sensors) and representations (explicit triangulated meshes, parametric body models, implicit surfaces, neural radiance fields, 3D Gaussian Splatting-based), Deep Reinforcement Learning and applications in physics-based behavior modeling.

Resources