1 code implementation • ICCV 2023 • Jungsoo Lee, Debasmit Das, Jaegul Choo, Sungha Choi
To be more specific, entropy minimization attempts to raise the confidence values of an individual sample's prediction, but individual confidence values may rise or fall due to the influence of signals from numerous other predictions (i. e., wisdom of crowds).
no code implementations • 31 Mar 2023 • Sunghyun Park, Sunghyo Chung, Jungsoo Lee, Jaegul Choo
However, STR models show a large performance degradation on languages with a numerous number of characters (e. g., Chinese and Korean), especially on characters that rarely appear due to the long-tailed distribution of characters in such languages.
1 code implementation • CVPR 2023 • Junha Song, Jungsoo Lee, In So Kweon, Sungha Choi
Second, our novel self-distilled regularization controls the output of the meta networks not to deviate significantly from the output of the frozen original networks, thereby preserving well-trained knowledge from the source domain.
1 code implementation • 27 Feb 2023 • Junwoo Park, Jungsoo Lee, Youngin Cho, Woncheol Shin, Dongmin Kim, Jaegul Choo, Edward Choi
Based on our findings, we propose a reweighting framework that down-weights the losses incurred by abrupt changes and up-weights those by normal states.
no code implementations • 14 Sep 2022 • Bum Chul Kwon, Jungsoo Lee, Chaeyeon Chung, Nyoungwoo Lee, Ho-Jin Choi, Jaegul Choo
We call the unwanted correlations "data biases," and the visual features causing data biases "bias factors."
no code implementations • 8 Jun 2022 • Jungsoo Lee, Juyoung Lee, Sanghun Jung, Jaegul Choo
Based on such issues, this paper 1) proposes an evaluation metric `Align-Conflict (AC) score' for the tuning criterion, 2) includes experimental settings with low bias severity and shows that they are yet to be explored, and 3) unifies the standardized experimental settings to promote fair comparisons between debiasing methods.
no code implementations • ICCV 2023 • Sanghun Jung, Jungsoo Lee, Nanhee Kim, Amirreza Shaban, Byron Boots, Jaegul Choo
That is, a model does not have a chance to learn test data in a class-discriminative manner, which was feasible in other adaptation tasks (\textit{e. g.,} unsupervised domain adaptation) via supervised losses on the source data.
no code implementations • 29 May 2022 • Jungsoo Lee, Jeonghoon Park, Daeyoung Kim, Juyoung Lee, Edward Choi, Jaegul Choo
$f_B$ is trained to focus on bias-aligned samples (i. e., overfitted to the bias) while $f_D$ is mainly trained with bias-conflicting samples by concentrating on samples which $f_B$ fails to learn, leading $f_D$ to be less susceptible to the dataset bias.
1 code implementation • 22 Oct 2021 • Jungsoo Lee, Jooyeol Yun, Sunghyun Park, Yonggyu Kim, Jaegul Choo
Despite the unprecedented improvement of face recognition, existing face recognition models still show considerably low performances in determining whether a pair of child and adult images belong to the same identity.
1 code implementation • ICCV 2021 • Sanghun Jung, Jungsoo Lee, Daehoon Gwak, Sungha Choi, Jaegul Choo
However, the distribution of max logits of each predicted class is significantly different from each other, which degrades the performance of identifying unexpected objects in urban-scene segmentation.
Ranked #4 on Anomaly Detection on Lost and Found
1 code implementation • NeurIPS 2021 • Jungsoo Lee, Eungyeup Kim, Juyoung Lee, Jihyeon Lee, Jaegul Choo
To this end, our method learns the disentangled representation of (1) the intrinsic attributes (i. e., those inherently defining a certain class) and (2) bias attributes (i. e., peripheral attributes causing the bias), from a large number of bias-aligned samples, the bias attributes of which have strong correlation with the target variable.