Deep Internal Learning

Speaker: Michal Irani , The Weizmann Institute of Science

Date: Wednesday, May 01, 2019

Time: 12:00 PM to 1:00 PM

Public: Yes

Location: 32-D463 Star Conference Room

Event Type: Seminar

Room Description:

Host: Bill Freeman, MIT CSAIL

Contact: Lukas Murmann, lmurmann@csail.mit.edu

Relevant URL:

Speaker URL: http://www.weizmann.ac.il/math/irani/home

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu

Reminder Subject: TALK: Michal Irani, "Deep Internal Learning"

In this talk I will show how complex visual inference tasks can be performed using Deep-Learning, in a totally unsupervised way, by exploiting the internal redundancy inside a single image. The strong recurrence of information inside a single natural image provides powerful internal examples which suffice for self-supervision of CNNs, without any prior examples or training data. This new “Deep Internal Learning” paradigm gives rise to true “Zero-Shot Learning”. I will show the power of this approach to a variety of problems, including super-resolution, image-segmentation, transparent layer separation, blind image-dehazing, image-retargeting, and more.

Research Areas:
AI & Machine Learning, Graphics & Vision

Impact Areas:

This event is not part of a series.

Created by Lukas Murmann Email at Friday, April 19, 2019 at 10:20 AM.