no code implementations • 21 Mar 2024 • Junhyeong Cho, Kim Youwang, Hunmin Yang, Tae-Hyun Oh
One of the biggest challenges in single-view 3D shape reconstruction in the wild is the scarcity of <3D shape, 2D image>-paired data from real-world environments.
no code implementations • 4 Mar 2024 • Hyunwoo Ha, Oh Hyun-Bin, Kim Jun-Seong, Kwon Byung-Ki, Kim Sung-Bin, Linh-Tam Tran, Ji-Yun Kim, Sung-Ho Bae, Tae-Hyun Oh
Video motion magnification is a technique to capture and amplify subtle motion in a video that is invisible to the naked eye.
1 code implementation • 7 Feb 2024 • Hansam Cho, Jonghyun Lee, Seoung Bum Kim, Tae-Hyun Oh, Yonghyun Jeong
Text-guided diffusion models have become a popular tool in image synthesis, known for producing high-quality and diverse images.
no code implementations • 10 Jan 2024 • GeonU Kim, Kim Youwang, Tae-Hyun Oh
FPRF efficiently stylizes large-scale 3D scenes by introducing a style-decomposed 3D neural radiance field, which inherits AdaIN's feed-forward stylization machinery, supporting arbitrary style reference images.
no code implementations • 18 Dec 2023 • Kim Youwang, Tae-Hyun Oh, Gerard Pons-Moll
We present Paint-it, a text-driven high-fidelity texture map synthesis method for 3D meshes via neural re-parameterized texture optimization.
2 code implementations • 15 Dec 2023 • Lee Hyun, Kim Sung-Bin, Seungju Han, Youngjae Yu, Tae-Hyun Oh
We introduce this new task to explain why people laugh in a particular video and a dataset for this task.
no code implementations • 15 Dec 2023 • Kwon Byung-Ki, Oh Hyun-Bin, Kim Jun-Seong, Hyunwoo Ha, Tae-Hyun Oh
In this work, we focus on improving legibility by proposing a new concept, axial motion magnification, which magnifies decomposed motions along the user-specified direction.
no code implementations • 2 Nov 2023 • Kim Sung-Bin, Lee Hyun, Da Hye Hong, Suekyeong Nam, Janghoon Ju, Tae-Hyun Oh
Laughter is a unique expression, essential to affirmative social interactions of humans.
no code implementations • 4 Oct 2023 • Kim Youwang, Lee Hyun, Kim Sung-Bin, Suekyeong Nam, Janghoon Ju, Tae-Hyun Oh
We propose NeuFace, a 3D face mesh pseudo annotation method on videos via neural re-parameterized optimization.
no code implementations • ICCV 2023 • Arda Senocak, Hyeonggon Ryu, Junsik Kim, Tae-Hyun Oh, Hanspeter Pfister, Joon Son Chung
However, prior arts and existing benchmarks do not account for a more important aspect of the problem, cross-modal semantic understanding, which is essential for genuine sound source localization.
no code implementations • 18 Sep 2023 • Minkyung Kim, Jongmin Yu, Junsik Kim, Tae-Hyun Oh, Jun Kyun Choi
Therefore, it has been a common practice to learn normality under the assumption that anomalous data are absent in a training dataset, which we call normality assumption.
no code implementations • 14 Aug 2023 • Kwon Byung-Ki, Kim Sung-Bin, Tae-Hyun Oh
Recent work on dense optical flow has shown significant progress, primarily in a supervised learning manner requiring a large amount of labeled data.
no code implementations • 2 Aug 2023 • Moon Ye-Bin, Nam Hyeon-Woo, Wonseok Choi, Nayeong Kim, Suha Kwak, Tae-Hyun Oh
Data imbalance in training data often leads to biased predictions from trained models, which in turn causes ethical and social issues.
no code implementations • ICCV 2023 • Moon Ye-Bin, Jisoo Kim, Hongyeob Kim, Kilho Son, Tae-Hyun Oh
Given the hypothesis, TextManiA transfers pre-trained text representation obtained from a well-established large language encoder to a target visual feature space being learned.
1 code implementation • 26 May 2023 • Seongyeon Park, Bohyung Kim, Tae-Hyun Oh
With our framework, we show superior performance compared to baselines in zero-shot TTS and VC, achieving state-of-the-art performance.
no code implementations • CVPR 2023 • Kim Sung-Bin, Arda Senocak, Hyunwoo Ha, Andrew Owens, Tae-Hyun Oh
The key idea is to enrich the audio features with visual information by learning to align audio to visual latent space.
1 code implementation • 30 Mar 2023 • Minkyu Kim, Kim Sung-Bin, Tae-Hyun Oh
Audio captioning aims to generate text descriptions from environmental sounds.
1 code implementation • 28 Mar 2023 • Seongyeon Park, Myungseo Song, Bohyung Kim, Tae-Hyun Oh
We empirically demonstrate the effectiveness of our proposed method in low-resource language scenarios, achieving outstanding performance compared to competing methods.
no code implementations • 20 Feb 2023 • Moon Ye-Bin, Dongmin Choi, Yongjin Kwon, Junsik Kim, Tae-Hyun Oh
We address a weakly-supervised low-shot instance segmentation, an annotation-efficient training method to deal with novel classes effectively.
1 code implementation • ICLR 2023 • Kwon Byung-Ki, Nam Hyeon-Woo, Ji-Yun Kim, Tae-Hyun Oh
Comprehensive studies of synthetic optical flow datasets have attempted to reveal what properties lead to accuracy improvement in learning-based optical flow estimation.
no code implementations • 1 Feb 2023 • Beichen Li, Bolei Deng, Wan Shou, Tae-Hyun Oh, Yuanming Hu, Yiyue Luo, Liang Shi, Wojciech Matusik
The conflict between stiffness and toughness is a fundamental problem in engineering materials design.
no code implementations • 26 Jan 2023 • Dong-Jin Kim, Tae-Hyun Oh, Jinsoo Choi, In So Kweon
We present a novel data-efficient semi-supervised framework to improve the generalization of image captioning models.
no code implementations • ICCV 2023 • Nam Hyeon-Woo, Kim Yu-Ji, Byeongho Heo, Doonyoon Han, Seong Joon Oh, Tae-Hyun Oh
We observe that the inclusion of CB reduces the degree of density in the original attention maps and increases both the capacity and generalizability of the ViT models.
1 code implementation • 14 Aug 2022 • Kim Jun-Seong, Kim Yu-Ji, Moon Ye-Bin, Tae-Hyun Oh
Our voxel-based volume rendering pipeline reconstructs HDR radiance fields with only multi-view LDR images taken from varying camera settings in an end-to-end manner and has a fast convergence speed.
1 code implementation • 27 Jul 2022 • Junhyeong Cho, Kim Youwang, Tae-Hyun Oh
Transformer encoder architectures have recently achieved state-of-the-art results on monocular 3D human mesh reconstruction, but they require a substantial number of parameters and expensive computations.
Ranked #3 on 3D Hand Pose Estimation on FreiHAND
1 code implementation • 9 Jun 2022 • Kim Youwang, Kim Ji-Yeon, Tae-Hyun Oh
Then, our novel zero-shot neural style optimization detailizes and texturizes the recommended mesh sequence to conform to the prompt in a temporally-consistent and pose-agnostic manner.
no code implementations • 12 Feb 2022 • Arda Senocak, Junsik Kim, Tae-Hyun Oh, Hyeonggon Ryu, DIngzeyu Li, In So Kweon
Human brain is continuously inundated with the multisensory information and their complex interactions coming from the outside world at any given moment.
no code implementations • 3 Nov 2021 • Kim Youwang, Kim Ji-Yeon, Kyungdon Joo, Tae-Hyun Oh
To make the unstable disjoint multi-task learning jointly trainable, we propose to exploit the morphological similarity between humans and animals, motivated by animal exercise where humans imitate animal poses.
no code implementations • 2 Oct 2021 • Yonghyun Jeong, Jooyoung Choi, Sungwon Kim, Youngmin Ro, Tae-Hyun Oh, Doyeon Kim, Heonseok Ha, Sungroh Yoon
In this work, we present Facial Identity Controllable GAN (FICGAN) for not only generating high-quality de-identified face images with ensured privacy protection, but also detailed controllability on attribute preservation for enhanced data utility.
no code implementations • 29 Sep 2021 • Dongmin Choi, Moon Ye-Bin, Junsik Kim, Tae-Hyun Oh
We propose the first weakly-supervised few-shot instance segmentation task and a frustratingly simple but strong baseline model, FoxInst.
1 code implementation • ICLR 2022 • Nam Hyeon-Woo, Moon Ye-Bin, Tae-Hyun Oh
We show that pFedPara outperforms competing personalized FL methods with more than three times fewer parameters.
no code implementations • ICCV 2021 • Donghyun Kim, Kuniaki Saito, Tae-Hyun Oh, Bryan A. Plummer, Stan Sclaroff, Kate Saenko
We present a two-stage pre-training approach that improves the generalization ability of standard single-domain pre-training.
1 code implementation • ICCV 2021 • Youmin Kim, Jinbae Park, YounHo Jang, Muhammad Ali, Tae-Hyun Oh, Sung-Ho Bae
In prevalent knowledge distillation, logits in most image recognition models are computed by global average pooling, then used to learn to encode the high-level and task-relevant knowledge.
Ranked #31 on Knowledge Distillation on ImageNet
1 code implementation • 8 Oct 2020 • Dong-Jin Kim, Tae-Hyun Oh, Jinsoo Choi, In So Kweon
To this end, we propose the multi-task triple-stream network (MTTSNet) which consists of three recurrent units responsible for each POS which is trained by jointly predicting the correct captions and POS for each word.
no code implementations • CVPR 2021 • Mallikarjun B R., Ayush Tewari, Tae-Hyun Oh, Tim Weyrich, Bernd Bickel, Hans-Peter Seidel, Hanspeter Pfister, Wojciech Matusik, Mohamed Elgharib, Christian Theobalt
The reflectance field of a face describes the reflectance properties responsible for complex lighting effects including diffuse, specular, inter-reflection and self shadowing.
no code implementations • 18 Mar 2020 • Donghyun Kim, Kuniaki Saito, Tae-Hyun Oh, Bryan A. Plummer, Stan Sclaroff, Kate Saenko
We show that when labeled source examples are limited, existing methods often fail to learn discriminative features applicable for both source and target domains.
1 code implementation • CVPR 2020 • Ruohan Gao, Tae-Hyun Oh, Kristen Grauman, Lorenzo Torresani
In the face of the video data deluge, today's expensive clip-level classifiers are increasingly impractical.
Ranked #8 on Action Recognition on ActivityNet
1 code implementation • 20 Nov 2019 • Arda Senocak, Tae-Hyun Oh, Junsik Kim, Ming-Hsuan Yang, In So Kweon
Visual events are usually accompanied by sounds in our daily lives.
no code implementations • 16 Sep 2019 • Seokju Lee, Junsik Kim, Tae-Hyun Oh, Yongseop Jeong, Donggeun Yoo, Stephen Lin, In So Kweon
We postulate that success on this task requires the network to learn semantic and geometric knowledge in the ego-centric view.
no code implementations • IJCNLP 2019 • Dong-Jin Kim, Jinsoo Choi, Tae-Hyun Oh, In So Kweon
To this end, our proposed semi-supervised learning method assigns pseudo-labels to unpaired samples via Generative Adversarial Networks to learn the joint distribution of image and caption.
3 code implementations • CVPR 2019 • Tae-Hyun Oh, Tali Dekel, Changil Kim, Inbar Mosseri, William T. Freeman, Michael Rubinstein, Wojciech Matusik
How much can we infer about a person's looks from the way they speak?
1 code implementation • CVPR 2019 • Junsik Kim, Tae-Hyun Oh, Seokju Lee, Fei Pan, In So Kweon
We take an approach to learn a generalizable embedding space for novel tasks.
1 code implementation • CVPR 2019 • Dong-Jin Kim, Jinsoo Choi, Tae-Hyun Oh, In So Kweon
Our goal in this work is to train an image captioning model that generates more dense and informative captions.
1 code implementation • 7 Feb 2019 • Alexandre Kaspar, Tae-Hyun Oh, Liane Makatura, Petr Kellnhofer, Jacqueline Aslarus, Wojciech Matusik
Motivated by the recent potential of mass customization brought by whole-garment knitting machines, we introduce the new problem of automatic machine instruction generation using a single image of the desired physical product, which we apply to machine knitting.
no code implementations • 27 Nov 2018 • Suwon Shon, Tae-Hyun Oh, James Glass
In this paper, we present a multi-modal online person verification system using both speech and visual signals.
no code implementations • CVPR 2018 • Kyungdon Joo, Tae-Hyun Oh, In So Kweon, Jean-Charles Bazin
In this work, we describe man-made structures via an appropriate structure assumption, called Atlanta world, which contains a vertical direction (typically the gravity direction) and a set of horizontal directions orthogonal to the vertical direction.
1 code implementation • 15 May 2018 • Changil Kim, Hijung Valentina Shin, Tae-Hyun Oh, Alexandre Kaspar, Mohamed Elgharib, Wojciech Matusik
We computationally model the overlapping information between faces and voices and show that the learned cross-modal representation contains enough information to identify matching faces and voices with performance similar to that of humans.
2 code implementations • ECCV 2018 • Tae-Hyun Oh, Ronnachai Jaroensri, Changil Kim, Mohamed Elgharib, Frédo Durand, William T. Freeman, Wojciech Matusik
We show that the learned filters achieve high-quality results on real videos, with less ringing artifacts and better noise characteristics than previous methods.
no code implementations • CVPR 2018 • Arda Senocak, Tae-Hyun Oh, Junsik Kim, Ming-Hsuan Yang, In So Kweon
We show that even with a few supervision, false conclusion is able to be corrected and the source of sound in a visual scene can be localized effectively.
no code implementations • 14 Feb 2018 • Dong-Jin Kim, Jinsoo Choi, Tae-Hyun Oh, Youngjin Yoon, In So Kweon
Human behavior understanding is arguably one of the most important mid-level components in artificial intelligence.
no code implementations • 5 Dec 2017 • Junsik Kim, Seokju Lee, Tae-Hyun Oh, In So Kweon
Recent advances in visual recognition show overarching success by virtue of large amounts of supervised data.
no code implementations • 24 Aug 2017 • Inwook Shim, Tae-Hyun Oh, Joon-Young Lee, Jinwook Choi, Dong-Geol Choi, In So Kweon
We introduce a novel method to automatically adjust camera exposure for image processing and computer vision applications on mobile robot platforms.
no code implementations • ICCV 2017 • Tae-Hyun Oh, Kyungdon Joo, Neel Joshi, Baoyuan Wang, In So Kweon, Sing Bing Kang
Cinemagraphs are a compelling way to convey dynamic aspects of a scene.
3 code implementations • ICCV 2017 • Donghyeon Cho, Jinsun Park, Tae-Hyun Oh, Yu-Wing Tai, In So Kweon
Our method implicitly learns an attention map, which leads to a content-aware shift map for image retargeting.
no code implementations • 6 Feb 2017 • Jinsoo Choi, Tae-Hyun Oh, In So Kweon
Despite the challenging baselines, our method still manages to show comparable or even exceeding performance.
no code implementations • NeurIPS 2016 • Tae-Hyun Oh, Yasuyuki Matsushita, In Kweon, David Wipf
Commonly used in many applications, robust PCA represents an algorithmic attempt to reduce the sensitivity of classical PCA to outliers.
no code implementations • CVPR 2016 • Kyungdon Joo, Tae-Hyun Oh, Junsik Kim, In So Kweon
Given a set of surface normals, we pose a Manhattan Frame (MF) estimation problem as a consensus set maximization that maximizes the number of inliers over the rotation search space.
no code implementations • CVPR 2016 • Jinsoo Choi, Tae-Hyun Oh, In So Kweon
Inspired by plot analysis of written stories, our method generates a sequence of video clips ordered in such a way that it reflects plot dynamics and content coherency.
no code implementations • 12 May 2016 • Kyungdon Joo, Tae-Hyun Oh, Junsik Kim, In So Kweon
Most man-made environments, such as urban and indoor scenes, consist of a set of parallel and orthogonal planar structures.
no code implementations • 12 Jan 2016 • Jinsoo Choi, Tae-Hyun Oh, In So Kweon
Photo collections and its applications today attempt to reflect user interactions in various forms.
no code implementations • 7 Dec 2015 • Tae-Hyun Oh, Yasuyuki Matsushita, In So Kweon, David Wipf
Commonly used in computer vision and other applications, robust PCA represents an algorithmic attempt to reduce the sensitivity of classical PCA to outliers.
no code implementations • 1 Sep 2015 • Tae-Hyun Oh, Yasuyuki Matsushita, Yu-Wing Tai, In So Kweon
The problems related to NNM, or WNNM, can be solved iteratively by applying a closed-form proximal operator, called Singular Value Thresholding (SVT), or Weighted SVT, but they suffer from high computational cost of Singular Value Decomposition (SVD) at each iteration.
no code implementations • CVPR 2015 • Tae-Hyun Oh, Yasuyuki Matsushita, Yu-Wing Tai, In So Kweon
The problems related to NNM (or WNNM) can be solved iteratively by applying a closed-form proximal operator, called Singular Value Thresholding (SVT) (or Weighted SVT), but they suffer from high computational cost to compute a Singular Value Decomposition (SVD) at each iteration.
no code implementations • 4 Mar 2015 • Tae-Hyun Oh, Yu-Wing Tai, Jean-Charles Bazin, Hyeongwoo Kim, In So Kweon
Robust Principal Component Analysis (RPCA) via rank minimization is a powerful tool for recovering underlying low-rank structure of clean data corrupted with sparse noise/outliers.