News

We discuss our findings at IEEE/CVF Conference on CVPR (Computer Vision and Pattern Recognition) 2021

19th – 25th June 2021 (virtual)

We are presenting 1 main conference paper, 2 (+1) workshop challenge winning methods and 1 invited workshop paper.
Main Conference Paper

Wednesday, June 23, 2021 6:00AM–8:30AM EDT | 7:00PM–9:30PM KST

We present a method to use the samples in episodic memory and an evaluation protocol for practical and realistic class incremental learning.

(Collaborated with NAVER AI)

Rainbow Memory: Continual Learning with a Memory of Diverse Samples

Jihwan Bang*, Heesu Kim*, Youngjoon Yoo, Jung-Woo Ha, Jonghyun Choi

CVPR 2021
PDF · SuppMat · Code
Workshop Challenge Winning Methods

Saturday, June 19 12:10PM EDT | Sunday, June 20 1:10AM KST

Event-based vision workshop

We present two methods in the workshop challenge. Here is the results.

(Collaborated with Prof. Kuk-Jin Yoon @ KAIST)

(1) EI-Stereo: (competition winner) using event and image information to estimate depth by stereo vision.

EI-Stereo: Stereo Depth Estimation from Events and Intensity Images

S. Mohammad Mostafavi I., Yeongwoo Nam, Kuk-Jin Yoon and Jonghyun Choi

Technical Report (presented at CVPR Workshop on event vision 2021)
PDF

(2) E-Stereo: (4th place) using event only on the same pipeline of EI-Stereo to estimate depth by stereo vision.

E-Stereo: Stereo Depth Estimation from Event Cameras

Yeongwoo Nam, S. Mohammad Mostafavi I., Kuk-Jin Yoon and Jonghyun Choi

Technical Report (presented at CVPR Workshop on event vision 2021)
PDF

Sunday, June 20 4:00PM PDT | Monday, June 21 8:00AM KST

2nd Embodied AI workshop

We present a method in the 'ALFRED challenge' in 'interaction challenge' section in the workshop. Here is the results.

(We win the 2nd place two years in a row! Collaborated with Dr. Roozbeh Mottaghi @ AI2)

ABP (Agent with the Big Picture): (2nd place)

Agent with the Big Picture: Perceiving Surroundings for Interactive Instruction Following

Byeonghwi Kim, Suvaansh Bhambri, Kunal Pratap Singh, Roozbeh Mottaghi, Jonghyun Choi

Technical Report (presented at CVPR Workshop on Embodied AI 2021)
PDF · Poster

Invited Workshop Paper

Friday, June 25 3:00PM-7:00PM EDT | Saturday, June 26 4:00AM-8:00AM KST

1st workshop on binary network

We present an extended method from our ECCV 2020 paper on Learning Architectures for Binary Networks, named BNAS v2

BNAS v2: A Summary with Empirical Improvements

Dahyun Kim, Kunal Pratap Singh, Jonghyun Choi

CVPR Workshop on binary network 2021
PDF · Poster · Talk Slide