Author name cluster

Minhyeok Lee

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

5 papers

1 author row

AAAI Conference 2026 Conference Paper

MonoCLUE: Object-Aware Clustering Enhances Monocular 3D Object Detection

Sunghun Yang
Minhyeok Lee
JungHo Lee
Sangyoun Lee

Monocular 3D object detection offers a cost-effective solution for autonomous driving, but it suffers from the ill-posed depth and a limited field of view. These constraints lead to the lack of geometric cues and reduced accuracy in occluded or truncated scenes. While recent approaches incorporate additional depth information to address geometric ambiguity, they overlook the importance of visual cues essential for robust object recognition. In this paper, we propose MonoCLUE that enhances monocular 3D detection by leveraging both local clustering and generalized scene memory of visual features. First, we perform K-means clustering on visual features to capture distinct object-level appearance visual parts (e.g., bonnet, car roof), which improves the detection of partially visible objects. The clustered features are then propagated across the entire region to capture objects with similar appearances. Second, we construct a generalized scene memory by aggregating clustered features across images, providing consistent appearance representations that generalize scenes. This improves the consistency of object-level features, enabling stable detection across varying environments. Lastly, we integrate both local cluster features and generalized scene memory into object queries, guiding attention toward informative regions in the feature map. Exploiting an unified local clustering and generalized scene memory strategy, MonoCLUE enables robust monocular 3D detection under occlusion and limited visibility. Our proposed model achieves state-of-the-art performance on the KITTI benchmark.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding

Minseok Kang
Minhyeok Lee
Minjung Kim
Donghyeong Kim
Sangyoun Lee

Video Temporal Grounding (VTG) aims to localize temporal segments in long, untrimmed videos that align with a given natural language query. This task typically comprises two subtasks: \textit{Moment Retrieval (MR)} and \textit{Highlight Detection (HD)}. While recent advances have been progressed by powerful pretrained vision-language models such as CLIP and InternVideo2, existing approaches commonly treat all text tokens uniformly during cross-modal attention, disregarding their distinct semantic roles. To validate the limitations of this approach, we conduct controlled experiments demonstrating that VTG models overly rely on [EOS]-driven global semantics while failing to effectively utilize word-level signals, which limits their ability to achieve fine-grained temporal alignment. Motivated by this limitation, we propose DualGround, a dual-branch architecture that explicitly separates global and local semantics by routing the [EOS] token through a sentence-level path and clustering word tokens into phrase-level units for localized grounding. Our method introduces (1) token-role-aware cross modal interaction strategies that align video features with sentence-level and phrase-level semantics in a structurally disentangled manner, and (2) a joint modeling framework that not only improves global sentence-level alignment but also enhances fine-grained temporal grounding by leveraging structured phrase-aware context. This design allows the model to capture both coarse and localized semantics, enabling more expressive and context-aware video grounding. DualGround achieves state-of-the-art performance on both Moment Retrieval and Highlight Detection tasks across QVHighlights and Charades-STA benchmarks, demonstrating the effectiveness of disentangled semantic modeling in video-language alignment.

PDF Details

AAAI Conference 2025 Conference Paper

Video Diffusion Models Are Strong Video Inpainter

Minhyeok Lee
Suhwan Cho
Chajin Shin
JungHo Lee
Sunghun Yang
Sangyoun Lee

Propagation-based video inpainting using optical flow at the pixel or feature level has recently garnered significant attention. However, it has limitations such as the inaccuracy of optical flow prediction and the propagation of noise over time. These issues result in non-uniform noise and time consistency problems throughout the video, which are particularly pronounced when the removed area is large and involves substantial movement. To address these issues, we propose a novel First Frame Filling Video Diffusion Inpainting model (FFF-VDI). We design FFF-VDI inspired by the capabilities of pre-trained image-to-video diffusion models that can transform the first frame image into a highly natural video. To apply this to the video inpainting task, we propagate the noise latent information of future frames to fill the masked areas of the first frame's noise latent code. Next, we fine-tune the pre-trained image-to-video diffusion model to generate the inpainted video. The proposed model addresses the limitations of existing methods that rely on optical flow quality, producing much more natural and temporally consistent videos. This proposed approach is the first to effectively integrate image-to-video diffusion models into video inpainting tasks. Through various comparative experiments, we demonstrate that the proposed model can robustly handle diverse inpainting types with high quality.

PDF Details DOI

EAAI Journal 2023 Journal Article

Portfolio optimization using predictive auxiliary classifier generative adversarial networks

Jiwook Kim
Minhyeok Lee

In financial engineering, portfolio optimization has been of consistent interest. Portfolio optimization is a process of modulating asset distributions to maximize expected returns and minimize risks. Despite numerous studies on shallow learning models, they have shown limited success in analyzing the complex nature of massive stock data, a task where recent deep learning models excel. However, the deterministic nature of conventional deep learning models impedes their consideration of portfolio risk due to an inherent lack of uncertainty quantification in their predictions. This paper proposes a novel portfolio weighting strategy, incorporating both risk and return considerations within a deep learning framework. We propose the Predictive Auxiliary Classifier Generative Adversarial Networks (PredACGAN), a probabilistic deep learning model, to measure prediction uncertainty. The PredACGAN generator leverages latent vectors and historical stock prices to predict future returns. The model synthesizes predictive distributions from various latent vectors and past prices. The associated risk is produced via the entropy of these distributions, facilitating portfolio optimization through both return and risk considerations. The proposed algorithm removes high-risk assets from the investment universe at rebalancing moments, enabling PredACGAN to optimize portfolios considering both return and risk. We evaluated PredACGAN and the accompanying algorithm with S&P 500 stocks from 1990 to 2020, with portfolios rebalanced monthly based on PredACGAN predictions and risk measures. The PredACGAN portfolios yielded 9. 123% annual returns and a 1. 054 Sharpe ratio, outperforming a risk-agnostic portfolio yielding 1. 024% annual returns and a 0. 236 Sharpe ratio. The PredACGAN portfolio also exhibited lower maximum drawdowns, highlighting its effectiveness.

Details DOI

EAAI Journal 2022 Journal Article

Inverse design of nanophotonic devices using generative adversarial networks

Wonsuk Kim
Soojeong Kim
Minhyeok Lee
Junhee Seok

The efficient design of structures that exhibit desired properties is challenging across various engineering and scientific applications. Traditional methods employ experts in a specific domain to design new structures with desired properties. Then, simulations are performed for the designed structures to evaluate whether they show desired properties, and such a process is with until the structures exhibit desired properties. Advances in computing power and machine learning have made these simulations and optimizations faster, but challenges remain that the researchers must perform optimizations in each iteration, which generally takes time and cost. A new framework called inverse design has been studied to address the limitations. In inverse design, structures with desired properties can directly be constructed. In this work, as an inverse design framework, we introduce a controllable generative adversarial network (ControlGAN) based model to generate nanophotonic devices with user-defined properties. As a result, the proposed model outperforms other GAN-based models when the model is evaluated by producing structures with maximum transmittance at specific wavelengths. Specifically, the proposed model achieves a mean F1-score of 0. 357, corresponding to a 260% improvement compared to the second-best model. The proposed model for inverse design can accelerate device designs not only in the field of nanophotonics but also in other nanostructures.

Details DOI