Premier Computer Vision Conference Accepts 10 Georgia Tech Papers
From helping chair umpires make better line calls in professional tennis to teaching robots to “see”, the field of computer vision continues to expand and become a part of people’s everyday lives. A subfield of artificial intelligence, computer vision teaches computers to understand and interpret the visual world through photos or videos.
The International Conference on Computer Vision (ICCV) takes place from Oct. 27 to Nov. 2 and brings together researchers from Georgia Tech and around the world to discuss recent breakthroughs and research in the field. Researchers in the Machine Learning Center at Georgia Tech (ML@GT) have ten accepted papers in the conference.
School of Interactive Computing (IC) and ML@GT associate professor Devi Parikh leads with seven research papers. Her work spans from using artificial intelligence (AI) to help people make more stylish outfit choices to embodied visual recognition.
IC assistant professor Judy Hoffman and professor James Rehg are 2019 area chairs.
“As the computer vision field continues to expand and create novel ideas, conferences like ICCV become increasingly important. There was a lot of impressive work submitted to the conference this year. With computer vision being one of ML@GT’s strongest areas, I’m thrilled to see the center’s presence in this premier conference,” said Hoffman.
Other work from Georgia Tech includes papers on lessening the need for additional annotation in videos, making vision and language models more grounded, and agents learning to move to better perceive objects.
"Having a paper accepted, especially as an oral presentation, especially in a top conference gives me lots of confidence and encouragement for my Ph.D. research. I can't wait to attend ICCV to share my work, talk with other talented people, and learn other interesting topics in both academic and industrial areas," said Min-Hung Chen, a sixth-year electrical and computer engineering Ph.D. student.
Organized by IEEE, ICCV is one of the premier international computer vision conferences and will take place at the COEX Convention Center in Seoul, South Korea.
The International Conference on Computer Vision (ICCV) takes place from Oct. 27 to Nov. 2 and brings together researchers from Georgia Tech and around the world to discuss recent breakthroughs and research in the field. Researchers in the Machine Learning Center at Georgia Tech (ML@GT) have ten accepted papers in the conference.
School of Interactive Computing (IC) and ML@GT associate professor Devi Parikh leads with seven research papers. Her work spans from using artificial intelligence (AI) to help people make more stylish outfit choices to embodied visual recognition.
IC assistant professor Judy Hoffman and professor James Rehg are 2019 area chairs.
“As the computer vision field continues to expand and create novel ideas, conferences like ICCV become increasingly important. There was a lot of impressive work submitted to the conference this year. With computer vision being one of ML@GT’s strongest areas, I’m thrilled to see the center’s presence in this premier conference,” said Hoffman.
Other work from Georgia Tech includes papers on lessening the need for additional annotation in videos, making vision and language models more grounded, and agents learning to move to better perceive objects.
"Having a paper accepted, especially as an oral presentation, especially in a top conference gives me lots of confidence and encouragement for my Ph.D. research. I can't wait to attend ICCV to share my work, talk with other talented people, and learn other interesting topics in both academic and industrial areas," said Min-Hung Chen, a sixth-year electrical and computer engineering Ph.D. student.
Organized by IEEE, ICCV is one of the premier international computer vision conferences and will take place at the COEX Convention Center in Seoul, South Korea.
From Annotation to Movement, Our Computer Vision Research Is Tackling It All
We know it's a challenge to make it to every presentation at a conference. If you're in a time crunch and have to miss one of our presentations (shame!), our students have broken down their research in the blogs below.
Overcoming Large-scale Annotation Requirements for Understanding Videos in the Wild
Create Annotating videos is time-consuming, but needed for training algorithms. This work explores a new method to reduce the amount of annotations needed. Catch the oral presentation of this work at 3:05 or at poster #36 on Thursday, October 31.
Embodied Amodal Recognition: Learning to Move to Perceive Objects
This poster details new work on embodied agents learning to move around an area in order to better perceive and understand objects. Check out this work at poster #77 on Tuesday, October 29 at 3:30 PM.
The 2019 Must-Have In Your Closet? AI.
Imagine being able to instantly become more stylish. Researchers from the University of Texas at Austin, Georgia Tech, and Facebook are working on a way to help people dress better by training an artificial intelligence agent to suggest minimal updates like tucking in a shirt or adding a belt.
Learn more about this exciting project at 10:30 on Thursday, October 31 at Poster 63.
Fashion++ in the News
- Temporal Attentive Alignment for Large-Scale Video Domain Adaptation by Min-Hung Chen, Zsolt Kira, Ghassan AlRegib, Jaekwon Yoo, Ruxin Chen, Jian Zheng
- *Best Paper Nominee* Habitat: A Platform for Embodied AI Research by Manolis Savva, Abhishek Kadian, Oleksandr Maksymets, Yili Zhao, Erik Wijmans, Bhavana Jain, Julian Straub, Jia Liu, Vladlen Koltun, Jitendra Malik, Devi Parikh, Dhruv Batra
-
SplitNet: Sim2Sim and Task2Task Transfer for Embodied Visual Navigation by
Daniel Gordon, Abhishek Kadian, Devi Parikh, Judy Hoffman, Dhruv Batra - Fashion++: Minimal Edits for Outfit Improvement by Wei-Lin Hsiao, Isay Katsman, Chao-Yuan Wu, Devi Parikh, Kristen Grauman
- Embodied Visual Recognition by Jianwei Yang*, Zhile Ren*, Mingze Xu, Xinlei Chen, David Crandall, Devi Parikh, Dhruv Batra
- Sequential Latent Spaces for Modeling the Intention During Diverse Image Captioning by Jyoti Aneja, Harsh Agrawal, Dhruv Batra, Alexander Schwing
- Floors are Flat: Leveraging Semantics for Real-Time Surface Normal Prediction by Steven Hickson, Karthik Raveendran, Alireza Fathi, Kevin Murphy, and Irfan Essa