no code implementations • ECCV 2020 • Sangeek Hyun, Jae-Pil Heo
As is well known, single image super-resolution (SR) is an ill-posed problem where multiple high resolution (HR) images can be matched to one low resolution (LR) image due to the difference of their representation capabilities.
1 code implementation • 20 Mar 2024 • Jiwoo Chung, Sangeek Hyun, Sang-Heon Shim, Jae-Pil Heo
Specifically, by assessing channel importance based on their sensitivities to latent vector perturbations, our method enhances the diversity of samples in the compressed model.
1 code implementation • 29 Dec 2023 • MinKyu Lee, Jae-Pil Heo
To tackle this issue, we propose a novel optimization method that can effectively remove the inherent noise term in the early steps of vanilla training by estimating the optimal centroid and directly optimizing toward the estimation.
1 code implementation • 27 Dec 2023 • Seunggu Kang, WonJun Moon, Euiyeon Kim, Jae-Pil Heo
Zero-Shot Object Counting (ZSOC) aims to count referred instances of arbitrary classes in a query image without human-annotated exemplars.
Ranked #2 on Zero-Shot Counting on FSC147
1 code implementation • 26 Dec 2023 • Sang-Heon Shim, Jiwoo Chung, Jae-Pil Heo
In this paper, we first investigate a visual quality degradation problem observed in recent high-resolution virtual try-on approach.
1 code implementation • 26 Dec 2023 • Suho Park, SuBeen Lee, Sangeek Hyun, Hyun Seok Seong, Jae-Pil Heo
Based on these two scores, we define a query background relevant score that captures the similarity between the backgrounds of the query and the support, and utilize it to scale support background features to adaptively restrict the impact of disruptive support backgrounds.
1 code implementation • 11 Dec 2023 • Jiwoo Chung, Sangeek Hyun, Jae-Pil Heo
Despite the impressive generative capabilities of diffusion models, existing diffusion model-based style transfer methods require inference-stage optimization (e. g. fine-tuning or textual inversion of style) which is time-consuming, or fails to leverage the generative ability of large-scale diffusion models.
2 code implementations • 15 Nov 2023 • WonJun Moon, Sangeek Hyun, SuBeen Lee, Jae-Pil Heo
Dummy tokens conditioned by text query take portions of the attention weights, preventing irrelevant video clips from being represented by the text query.
Ranked #1 on Highlight Detection on TvSum
no code implementations • ICCV 2023 • JiHwan Kim, Miso Lee, Jae-Pil Heo
In this paper, we point out the problem in the self-attention of DETR for TAD; the attention modules focus on a few key elements, called temporal collapse problem.
1 code implementation • 28 Jul 2023 • SuBeen Lee, WonJun Moon, Hyun Seok Seong, Jae-Pil Heo
While TDM influences high-level feature maps by task-adaptive calibration of channel-wise importance, we further introduce Instance Attention Module (IAM) operating in intermediate layers of feature extractors to instance-wisely highlight object-relevant channels, by extending QAM.
1 code implementation • CVPR 2023 • Hyun Seok Seong, WonJun Moon, SuBeen Lee, Jae-Pil Heo
Specifically, we add the loss propagating to local hidden positives, semantically similar nearby patches, in proportion to the predefined similarity scores.
Ranked #2 on Unsupervised Semantic Segmentation on Potsdam-3
1 code implementation • CVPR 2023 • WonJun Moon, Sangeek Hyun, Sanguk Park, Dongchan Park, Jae-Pil Heo
As we observe the insignificant role of a given query in transformer architectures, our encoding module starts with cross-attention layers to explicitly inject the context of text query into video representation.
Ranked #2 on Highlight Detection on TvSum
no code implementations • CVPR 2023 • Haechan Noh, Sangeek Hyun, Woojin Jeong, Hanshin Lim, Jae-Pil Heo
The inverted index is a widely used data structure to avoid the infeasible exhaustive search.
1 code implementation • 24 Nov 2022 • WonJun Moon, Hyun Seok Seong, Jae-Pil Heo
A dramatic increase in real-world video volume with extremely diverse and emerging topics naturally forms a long-tailed video distribution in terms of their categories, and it spotlights the need for Video Long-Tailed Recognition (VLTR).
1 code implementation • 20 Jul 2022 • WonJun Moon, Ji-Hwan Kim, Jae-Pil Heo
Our exhaustive experiments validate the merits of LoRot as a pretext task tailored for supervised learning in terms of robustness and generalization capability.
Ranked #9 on Data Augmentation on ImageNet
1 code implementation • 20 Jul 2022 • WonJun Moon, Junho Park, Hyun Seok Seong, Cheol-Ho Cho, Jae-Pil Heo
Furthermore, moderate- and easy-difficulty samples are also yielded by our modified GAN and Copycat, respectively.
1 code implementation • CVPR 2022 • SuBeen Lee, WonJun Moon, Jae-Pil Heo
Specifically, TDM learns task-specific channel weights based on two novel components: Support Attention Module (SAM) and Query Attention Module (QAM).
Ranked #9 on Few-Shot Image Classification on CUB 200 5-way 5-shot (using extra training data)
no code implementations • CVPR 2022 • Sang-Heon Shim, Sangeek Hyun, DaeHyun Bae, Jae-Pil Heo
To address this, we propose a novel attention module, Local Attention Pyramid (LAP) module tailored for scene image synthesis, that encourages GANs to generate diverse object classes in a high quality by explicit spread of high attention scores to local regions, since objects in scene images are scattered over the entire images.
1 code implementation • IEEE Access 2021 • Jaihyun Lew, Euiyeon Kim, Jae-Pil Heo
Furthermore, based on this assumption, there have been attempts to estimate the blur kernel of a given LR image, since correct kernel priors are known to be helpful in super-resolution.
Ranked #5 on Blind Super-Resolution on DIV2KRK - 4x upscaling
no code implementations • CVPR 2021 • Sangeek Hyun, JiHwan Kim, Jae-Pil Heo
The proposed tasks enable the discriminators to learn representations of appearance and temporal context, and force the generator to synthesize videos with consistent appearance and natural flow of motions.
no code implementations • ICCV 2021 • Haechan Noh, TaeHo Kim, Jae-Pil Heo
To address the raised question, we suggest a joint optimization of the coarse and fine quantizers by substituting the original objective of the coarse quantizer to end-to-end quantization distortion.
no code implementations • ICCV 2021 • Won Young Jhoo, Jae-Pil Heo
In such a case, we can capture the domain shift between the source domain and target domain from an unseen task and transfer it to the task of interest, which is known as zero-shot domain adaptation (ZSDA).
no code implementations • CVPR 2016 • Jae-Pil Heo, Zhe Lin, Xiaohui Shen, Jonathan Brandt, Sung-Eui Yoon
We have tested the proposed method with the inverted index and multi-index on a diverse set of benchmarks including up to one billion data points with varying dimensions, and found that our method robustly improves the accuracy of shortlists (up to 127% relatively higher) over the state-of-the-art techniques with a comparable or even faster computational cost.
no code implementations • CVPR 2014 • Jae-Pil Heo, Zhe Lin, Sung-Eui Yoon
This result is achieved mainly because our method accurately estimates distances between two data points with the new binary codes and distance metric.