קולוקוויום בביה"ס למדעי המחשב - Exploring Image Collections

Hadar Elor
02 ביוני 2019, 11:00 
בניין שרייבר, חדר 006 
קולוקוויום במדעי המחשב

Nowadays, every single day billions of photos are uploaded to the internet. As a consequence, we are currently experiencing a dramatic growth in the size of searchable digital image repositories. Thousands — sometimes millions — of images are at a user’s finger tips, for every possible query phrase. Text-based queries through search engines such as Google, Bing, and Flickr, allow combing through this sea of data and enable extracting large topical but otherwise unstructured wild image collections.

In this talk, I will present recent techniques to explore and analyze large image collections. I will introduce a technique to traverse these collections in an as-smooth-as-possible manner and demonstrate how these smooth sequences can be manipulated to allow for a data-driven morphing technique. I will also present a technique to automatically manipulate facial expression in images given a single image of a target face. Finally, I will present a clustering-driven deep embedding technique to group large collections in a non-parametric setting, where the number of clusters is unknown.

אוניברסיטת תל אביב עושה כל מאמץ לכבד זכויות יוצרים. אם בבעלותך זכויות יוצרים בתכנים שנמצאים פה ו/או השימוש
שנעשה בתכנים אלה לדעתך מפר זכויות, נא לפנות בהקדם לכתובת שכאן >>