publications

2025

1. CVPR

   

   Exploiting Deblurring Networks for Radiance Fields

   Haeyun Choi, Heemin Yang, Janghyuk Han, and Sunghyun Cho

   In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025

   Outstanding Paper Award at IPIU 2025

   Abs arXiv Dataset Code Website

   In this paper, we propose DeepDeblurRF, a novel radiance field deblurring approach that can synthesize high-quality novel views from blurred training views with significantly reduced training time. DeepDeblurRF leverages deep neural network (DNN)-based deblurring modules to enjoy their deblurring performance and computational efficiency. To effectively combine DNN-based deblurring and radiance f ield construction, we propose a novel radiance field (RF)guided deblurring and an iterative framework that performs RF-guided deblurring and radiance field construction in an alternating manner. Moreover, DeepDeblurRF is compatible with various scene representations, such as voxel grids and 3D Gaussians, expanding its applicability. We also present BlurRF-Synth, the first large-scale synthetic dataset for training radiance field deblurring frameworks. We conduct extensive experiments on both camera motion blur and defocus blur, demonstrating that DeepDeblurRF achieves state-of-the-art novel-view synthesis quality with significantly reduced training time.

2024

1. IPIU

   

   Optical Flow-based Shakiness Detection in Dash-cams for Hit-and-Run Accident Analysis

   Haeyun Choi, and Sunghyun Cho

   In Workshop on Image Processing and Image Understanding (IPIU), 2024

   Abs PDF

   In this paper, we present a straightforward approach for detecting video shakiness in dash-cams and demonstrate its potential application in analyzing hit-and-run accidents. We use optical flow to calculate directional components of motion vectors for each key point, uniformly distributed across every frame. Subsequently, we leverage the trend of the standard deviation of these motion vectors’ directions to detect occurrences of shake in dash-cam footage. Since there is no publicly available dataset related to hit-and-run accidents, we collected a total of 63 dash-cam videos to evaluate the performance of our algorithm. The experimental results show that the proposed method achieves high detection performance across various types and levels of hit-and-run accidents even under low-light or nighttime conditions.

2023

1. arXiv

   

   AutoCycle-VC: Towards Bottleneck-Independent Zero-Shot Cross-Lingual Voice Conversion

   Haeyun Choi, Jio Gim, Yuho Lee, Youngin Kim, and Youngjoo Suh

   In arXiv preprint, 2023

   Abs arXiv

   This paper proposes a simple and robust zero-shot voice conversion system with a cycle structure and mel-spectrogram pre-processing. Previous works suffer from information loss and poor synthesis quality due to their reliance on a carefully designed bottleneck structure. Moreover, models relying solely on self-reconstruction loss struggled with reproducing different speakers’ voices. To address these issues, we suggested a cycle-consistency loss that considers conversion back and forth between target and source speakers. Additionally, stacked random-shuffled mel-spectrograms and a label smoothing method are utilized during speaker encoder training to extract a time-independent global speaker representation from speech, which is the key to a zero-shot conversion. Our model outperforms existing state-of-the-art results in both subjective and objective evaluations. Furthermore, it facilitates cross-lingual voice conversions and enhances the quality of synthesized speech.