no code implementations • 2 Feb 2024 • Bum Jun Kim, Sang Woo Kim
We discover that scale disequilibrium is caused by bilinear upsampling, which is supported by both theoretical and empirical evidence.
no code implementations • 7 Nov 2023 • Bum Jun Kim, Hyeonah Jang, Sang Woo Kim
The latest advances in deep learning have facilitated the development of highly accurate monocular depth estimation models.
no code implementations • 26 Jul 2023 • Bum Jun Kim, Hyeyeon Choi, Hyeonah Jang, Sang Woo Kim
However, fixed values of atrous rates are used for the ASPP module, which restricts the size of its field of view.
Ranked #1 on Retinal Vessel Segmentation on HRF (mIoU metric)
1 code implementation • 8 May 2023 • Bum Jun Kim, Hyeyeon Choi, Hyeonah Jang, Sang Woo Kim
Based on this observation, we propose explicitly adding a Gaussian attention bias that guides the positional embedding to have the corresponding pattern from the beginning of training.
1 code implementation • 13 Feb 2023 • Bum Jun Kim, Hyeyeon Choi, Hyeonah Jang, Donggeon Lee, Sang Woo Kim
In this study, we investigate the correct position to apply dropout.
no code implementations • 7 Feb 2023 • Bum Jun Kim, Hyeyeon Choi, Hyeonah Jang, Sang Woo Kim
First, we find that the number of groups influences the gradient behavior of the group normalization layer.
no code implementations • 15 May 2022 • Bum Jun Kim, Hyeyeon Choi, Hyeonah Jang, Dong Gu Lee, Wonseok Jeong, Sang Woo Kim
L2 regularization for weights in neural networks is widely used as a standard training trick.
Ranked #2 on Text Classification on GLUE SST2
no code implementations • 16 Nov 2021 • Bum Jun Kim, Hyeyeon Choi, Hyeonah Jang, Dong Gu Lee, Wonseok Jeong, Sang Woo Kim
We compared the robustness of CNN and ViT by assuming various image corruptions that may appear in practical vision tasks.
1 code implementation • 31 Aug 2021 • Bum Jun Kim, Hyeyeon Choi, Hyeonah Jang, Dong Gu Lee, Wonseok Jeong, Sang Woo Kim
First, we evaluated the size of the receptive field.
Ranked #5 on Fine-Grained Image Classification on Caltech-101
no code implementations • 15 Jan 2020 • Bum Jun Kim, Gyogwon Koo, Hyeyeon Choi, Sang Woo Kim
First, we propose Gaussian upsampling, an improved upsampling method that can reflect the characteristics of deep learning models.
no code implementations • 9 Oct 2018 • Sang Jun Lee, Sang Woo Kim, Wookyong Kwon, Gyogwon Koo, Jong Pil Yun
The main contribution of this paper is on selective distillation to improve the quality of the weakly annotated GTD.