Pinar Yanardag

Featuring a dress and a necklace generated by GANs -- made by human designers at Conference of the Future'19, Chile.

My principal research interest lies in developing deep learning approaches to enable human-AI collaboration, creativity and interpretability. My research is published at top computer science conferences such as NeurIPS, ICCV and KDD, and featured in mainstream media (e.g., The Washington Post, BBC, CNN) and magazines (e.g., Motherboard, Rolling Stone). For more information, see the Publications page.

Currently, I am leading a research lab at Bogazici University under the TUBITAK Outstanding Research Fellows program. Previously, I was a postdoctoral research associate at MIT Media Lab. I did my Ph.D. in Computer Science at Purdue, under the supervision of SVN Vishwanathan. During my Ph.D., I also spent time at Amazon (P13N team) and VMware. I received my Master’s degree from Bogazici University under the supervision of Ethem Alpaydin. During my graduate studies, I was supported by Fulbright PhD Fellowship and Google Anita Borg Scholarship.

I am also the founder of AI Fiction, a creative design studio specializing in artificial intelligence. Some of our work includes designing AI algorithms for HBO’s Westworld, for which I was a Creative Director nominee at 72nd Primetime Emmy Awards. I am also a fashion designer and co-founded GLITCH–the world’s first clothing line with fashion designs created by GANs and made in real life by human collaborators.

I regularly teach the course How to Generate (Almost) Anything which focuses on generative models and human-AI collaboration. I am also interested in teaching AI to fashion designers and taught a course at London College of Fashion called “AI & Fashion”. See Courses section for more information.

For a full CV, please see this PDF.

news

Jan 22, 2023	Deep Empathy is featured on New York Times!
Aug 15, 2022	One paper is accepted to ECCV’22, two papers are accepted to WACV’23!
Apr 1, 2022	Invited talk at CMU Robotics Institute, watch live or visit on campus!
Mar 15, 2022	Invited talk at MIT EECS Vision & Graphics Seminar!
Feb 1, 2022	CATLAB undergraduate researchers Cemre and Eylul won the best senior thesis award!

selected publications

ECCV

FairStyle: Debiasing StyleGAN2 with Style Channel Manipulations

Cemre Efe Karakas*, Alara Dirik*, Eylul Yalcinkaya, and Pinar Yanardag

European Conference on Computer Vision 2022

Abstract PDF Website

Recent advances in generative adversarial networks have shown that it is possible to generate high-resolution and hyperrealistic images. However, the images produced by GANs are only as fair and representative as the datasets on which they are trained. In this paper, we propose a method for directly modifying a pre-trained StyleGAN2 model that can be used to generate a balanced set of images with respect to one (e.g., eyeglasses) or more attributes (e.g., gender and eyeglasses). Our method takes advantage of the style space of the StyleGAN2 model to perform disentangled control of the target attributes to be debiased. Our method does not require training additional models and directly debiases the GAN model, paving the way for its use in various downstream applications. Our experiments show that our method successfully debiases the GAN model within a few minutes without compromising the quality of the generated images.
ICCV

LatentCLR: A Contrastive Learning Approach for Unsupervised Discovery of Interpretable Directions

Oguz Kaan Yuksel, Enis Simsar, Ezgi Gulperi Er, and Pinar Yanardag

International Conference on Computer Vision (ICCV) 2021

Abstract PDF Code

Recent research has shown great potential for finding interpretable directions in the latent spaces of pre-trained Generative Adversarial Networks (GANs). These directions provide controllable generation and support a wide range of semantic editing operations such as zoom or rotation. The discovery of such directions is often performed in a supervised or semi-supervised fashion and requires manual annotations, limiting their applications in practice. In comparison, unsupervised discovery enables finding subtle directions a priori hard to recognize. In this work, we propose a contrastive-learning-based approach for discovering semantic directions in the latent space of pretrained GANs in a self-supervised manner. Our approach finds semantically meaningful dimensions compatible with state-of-the-art methods.