Seil Na

Hello! I am a Research Scientist at Lunit Inc. Previously, I was a master’s student in Dept. of Computer Science and Engineering at Seoul National University, where I was advised by Prof. Gunhee Kim.

My research is on Computer Vision, Natural Language Processing and Machine Learning. I'm also interested in training neural networks on large-scale dataset.

Email  /  CV  /  Google Scholar  /  GitHub

News
Research
Learning Visual Context by Comparison
Minchul Kim*, Jongchan Park*, Seil Na, Chang Min Park, Donggeun Yoo (*: eqaul contribution)
ECCV, 2020 (spotlight)
paper / code / OpenReview / poster / bibtex

We propose Attent-and-Compare Module for chest X-ray abnomality detection task, whose key concept is capturing the difference between an object of interest and its corresponding context in image.

Discovery of Natural Language Concepts in Individual Units of CNNs
Seil Na, Yo Joong Choe, Dong-Hyun Lee, Gunhee Kim
ICLR, 2019
paper / code / OpenReview / poster / bibtex

In an attempt to understand the representations of deep convolutional networks trained on language tasks, we show that individual units are selectively responsive to specific morphemes, words, and phrases, rather than responding to arbitrary and uninterpretable patterns.

Encoding Video and Label Priors for Multi-label Video Classification on YouTube-8M dataset
Seil Na, Youngjae Yu, Sangho Lee, Jisung Kim, Gunhee Kim
CVPR Workshop on YouTube-8M Large-Scale Video Understanding, 2017
paper / code / bibtex

We propose a deep neural network model, which consists of four components: the frame encoder, the classification layer, the label processing layer, and the loss function. We ranked 8th out of 655 teams in Kaggle - Google Cloud & YouTube-8M Video Understanding Challenge.

A Read-Write Memory Network for Movie Story Understanding
Seil Na, Sangho Lee, Jisung Kim, Gunhee Kim
ICCV, 2017
paper / code / poster / bibtex

We propose a novel memory network model named Read-Write Memory Network (RWMN) to perform question and answering tasks for large-scale, multimodal movie story understanding.