Colloquium: Danna Gurari: Mixing Machines, Crowds, and Experts for Scalable Image Annotation



The ubiquitous use of cameras is yielding a rich abundance of visual information that can be leveraged for many valuable applications. However, the key bottlenecks for building these applications lie in unreliable annotation from algorithms and costly annotation by experts, especially at scale. Given the rise of crowdsourcing, I explore how we can utilize online crowds with algorithms and experts to better annotate images.

I will present research on demarcating objects in images (segmentation), a critical and time-consuming precursor to many downstream applications. A running theme in my work is to estimate and account for expert disagreement when designing and evaluating systems. The first part of my presentation is based on the observation that experts often disagree because tasks are difficult (i.e., complex object boundaries). I will discuss how to efficiently leverage crowd and algorithm efforts independently and jointly in order to optimize cost/quality trade-offs and produce segmentations comparable to those created by experts. The second part of my presentation is based on the observation that experts also disagree because tasks are ambiguous (i.e., which is the most prominent object in an image?). I will describe a system for predicting whether an image is ambiguous and so needs multiple annotations to capture the diversity of valid segmentations. In addition to analyzing everyday images, an emphasis of my work is on biomedical/medical images in order to contribute to research which targets society?s health problems.


Danna Gurari is currently a Postdoctoral Fellow at University of Texas at Austin under the supervision of Dr. Kristen Grauman. She completed her PhD at Boston University in the Image and Video Computing Group under the supervision of Dr. Margrit Betke. Her research interests span computer vision, human computation/crowdsourcing, medical/biomedical image analysis, and applied machine learning. In 2007-2010, Danna worked at Boulder Imaging building custom, high performance, multi-camera recording and analysis systems for military, industrial, and academic applications. From 2005-2007, she worked at Raytheon developing software for satellite systems. Danna earned her BS in Biomedical Engineering and MS in Computer Science from Washington University in St. Louis in 2005, with her thesis on ultrasound imaging. Danna was awarded the 2015 Researcher Excellence Award from the Boston University computer science department, 2014 Best Paper Award for Innovative Idea at MICCAI IMIC, and 2013 Best Paper Award at WACV.


8:15am to 9:30am


glqxz9283 sfy39587stf02 mnesdcuix8