BLISS Speaker Series Episode #19

The organizers are thrilled to feature Dr. Pierre Fernandez, Research Scientist at Meta AI, as the guest speaker for an engaging talk titled "Watermark Anything with Localized Messages". The presentation will run for approximately 45 minutes and will delve into state-of-the-art techniques in invisible image watermarking.

Following the talk, attendees are invited to stay for a social gathering—an excellent chance to connect with fellow AI enthusiasts, exchange ideas, ask questions, and enjoy complimentary drinks. Please note that doors will close at 7:15 PM, so arriving early is strongly encouraged.

RSVP is strictly required via Meetup to guarantee entry.

Meetup has recently increased promotion of its "Plus" program, but there is no obligation to purchase it. Both the platform and this event remain completely free.

Who Should Attend?
This event welcomes anyone interested in cutting-edge AI research, with a particular emphasis on students, PhD candidates, academic researchers, and professionals in machine learning and related fields.

Talk Abstract:
Invisible image watermarking involves embedding information into images in a way that is imperceptible to the human eye but remains retrievable—even after extensive editing. However, conventional watermarking methods often fail when the watermarked regions are small or localized, a common challenge in real-world applications involving altered or composite images.

In this talk, Dr. Fernandez will introduce an innovative approach: Watermark Anything (ICLR 2025), which reframes watermarking as a segmentation problem. He will discuss the motivation behind this perspective, the development and training of the model, the practical challenges encountered, and the outcomes of this research.

Speaker Bio:
Dr. Pierre Fernandez is a Research Scientist at FAIR Paris, the Fundamental AI Research lab at Meta. His work focuses on content protection and safety in machine learning, with a particular emphasis on watermarking generative models. He earned his PhD at FAIR and Inria Rennes, under the guidance of Matthijs Douze, Hervé Jégou, and Teddy Furon. Prior to his doctoral studies, he graduated from École Polytechnique, where he specialized in computer science and mathematics for visual computing.

Technische Universität Berlin

Straße des 17.Juni 135, 10623 Berlin

Zur Registrierung