Egocentric Computer Vision: Perception and Understanding from the First-Person View 2026

This seminar explores the rapidly evolving field of egocentric computer vision, focusing on how first-person visual data, optionally combined with on-body sensors, enables understanding of human behavior and interaction with the surrounding world. The course surveys current research in egocentric perception, covering high-level topics such as pose estimation, activity recognition, scene understanding, 3D reconstruction, and generative models from egocentric viewpoints. Through the study and presentation of recent research papers, students develop the ability to critically assess methodologies, interpret experimental results, and communicate scientific contributions. The seminar emphasizes discussion and collaborative analysis, fostering a research-oriented learning environment. Each session features a student-led presentation of a selected paper followed by a group discussion. Active participation is expected: students prepare assigned readings, present papers, and contribute constructively to peer discussions.

Overview

Seminar
263-5909-00 S Egocentric Computer Vision: Perception and Understanding from the First-Person View
Lecturers
Christian Holz, Dominik Hollidt, Jiaxi Jiang
Communication
Please address all questions (on content, organization, etc.) on Moodle
link to Egocentric Computer Vision Moodle 2026
Lecture
ETH seminar room STD G1.
Wednesdays, 4–6pm
first seminar: 18.02.2026
last seminar: 27.05.2026
Credits
2 ECTS
Materials
slides and assignments