Two School of Interactive Computing Ph.D. students, Abhishek Das and Patsorn Sangkloy, were selected as 2018 Adobe Research Fellows.
A total of eight were selected by Adobe Research from a highly competitive field of applicants in computer graphics, computer vision, human-computer interaction, machine learning, visualization, audio, natural language processing, and programming languages.
The fellowship awards its selections $10,000 funding and an optional opportunity to intern at Adobe for summer 2018. They also are given a subscription to Adobe’s Creative Cloud service.
Sangkloy’s research focuses on providing users with images that match with their intent based on line drawing sketch. This entails a handful of options:
In sketch-based image retrieval, the research suggests the ability to retrieve images based on association with a sketched query. In sketch-based image synthesis, an entirely new image is generated directly from the sketch input, where editing the sketch produces different generated results.
Lastly, she and her collaborators are currently working with Adobe researchers to develop an image retrieval pipeline specifically designed for human images. Since drawing an entire body of a human is no easy feat, even for a professional artist, they explore the option of combining stick figures and/or sparse sketch outlines.
She is advised by Associate Professor James Hays.
Das’ research focuses on deep learning and its applications in building agents that can see, think, talk, and act. To this end, he was awarded the fellowship for his ongoing work on the following:
- Agents that can hold free-form conversations about visual content and algorithms to train these agents without exhaustively collecting human-annotated datasets, via simulation or self-play;
- Algorithms to make these agents more interpretable;
- Agents that can execute commands, including navigating and actively perceiving their environments to accomplish a task, like question-answering.
This was Das’ second fellowship selection after also earning one from Snap Research for a similar proposal. He is advised by Assistant Professor Dhruv Batra.