Kim Jun-Seong
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Kim Jun-Seong has a total of 39 co-authors. Here are the top co-authors by number of shared publications:
- Byung-Sung Kim
College of Information and Communication Engineering, Sungkyunkwan University, Suwon, South Korea
7 shared publications - Tae-Hyun Oh
KAIST
5 shared publications - Jae-Hyun Park
Department of Electronic, Electrical, and Computer Engineering Sungkyunkwan University Suwon Korea
4 shared publications - Reem Song
College of Information and Communication Engineering, Sungkyunkwan University, Suwon, South Korea
4 shared publications - GeonU Kim
KAIST
2 shared publications - Kim Yu-Ji
POSTECH
2 shared publications
Latest Publications
- 11 September 2026
- Published by Springer Nature
- p. 133-151
SplatReasoner: Enhancing Embodied Reasoning and Grounding by Novel View Synthesis
- 11 September 2026
- Published by Springer Nature
- p. 133-151
SplatReasoner: Enhancing Embodied Reasoning and Grounding by Novel View Synthesis
- 11 September 2026
- Published by Springer Nature
- p. 300-319
SA-ResGS: Self-augmented Residual 3D Gaussian Splatting for Next Best View Selection
- 11 September 2026
- Published by Springer Nature
- p. 300-319
SA-ResGS: Self-augmented Residual 3D Gaussian Splatting for Next Best View Selection
- Open Access
- 11 April 2025
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 39 (7) , 7167-7175
SoundBrush: Sound as a Brush for Visual Scene Editing
We propose SoundBrush, a model that uses sound as a brush to edit and manipulate visual scenes. We extend the generative capabilities of the Latent Diffusion Model (LDM) to incorporate audio information for editing visual scenes. Inspired by existing image-editing works, we frame this task as a supervised learning problem and leverage various off-the-shelf models to construct a sound-paired visual scene editing dataset for training. This richly generated dataset enables SoundBrush to learn to map audio features into the textual space of the LDM, allowing for visual scene editing guided by diverse in-the-wild sound. Unlike existing methods, SoundBrush can accurately manipulate the overall scenery or even insert sounding objects to best match the input sound semantics while preserving the original content. Furthermore, by integrating with novel view synthesis techniques, our framework can be extended to edit 3D scenes, facilitating sound-driven 3D scene manipulation.
- 11 April 2025
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 39 (7) , 7167-7175
Open AccessSoundBrush: Sound as a Brush for Visual Scene Editing
We propose SoundBrush, a model that uses sound as a brush to edit and manipulate visual scenes. We extend the generative capabilities of the Latent Diffusion Model (LDM) to incorporate audio information for editing visual scenes. Inspired by existing image-editing works, we frame this task as a supervised learning problem and leverage various off-the-shelf models to construct a sound-paired visual scene editing dataset for training. This richly generated dataset enables SoundBrush to learn to map audio features into the textual space of the LDM, allowing for visual scene editing guided by diverse in-the-wild sound. Unlike existing methods, SoundBrush can accurately manipulate the overall scenery or even insert sounding objects to best match the input sound semantics while preserving the original content. Furthermore, by integrating with novel view synthesis techniques, our framework can be extended to edit 3D scenes, facilitating sound-driven 3D scene manipulation.
- 31 October 2024
- Published by Springer Nature
- p. 179-195
Learning-based Axial Video Motion Magnification
- 31 October 2024
- Published by Springer Nature
- p. 179-195
Learning-based Axial Video Motion Magnification
- 13 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (11) , 10272-10279
Factorized Multi-Resolution HashGrid for Efficient Neural Radiance Fields: Execution on Edge-Devices
We introduce Fact-Hash, a novel parameter-encoding method for training on-device neural radiance fields. Neural Radiance Fields (NeRF) have proven pivotal in 3D representations, but their applications are limited due to large computational resources. On-device training can open large application fields, providing strength in communication limitations, privacy concerns, and fast adaptation to a frequently changing scene. However, challenges such as limited resources (GPU memory, storage, and power) impede their deployment. To handle this, we introduce Fact-Hash, a novel parameter-encoding merging Tensor Factorization and Hash-encoding techniques. This integration offers two benefits: the use of rich high-resolution features and the few-shot robustness. In Fact-Hash, we project 3D coordinates into multiple lower-dimensional forms (2D or 1D) before applying the hash function and then aggregate them into a single feature. Comparative evaluations against state-of-the-art methods demonstrate Fact-Hash's superior memory efficiency, preserving quality and rendering speed. Fact-Hash saves memory usage by over one-third while maintaining the PSNR values compared to previous encoding methods. The on-device experiment validates the superiority of Fact-Hash compared to alternative positional encoding methods in computational efficiency and energy consumption. These findings highlight Fact-Hash as a promising solution to improve feature grid representation, address memory constraints, and improve quality in various applications.
- 13 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (11) , 10272-10279
Factorized Multi-Resolution HashGrid for Efficient Neural Radiance Fields: Execution on Edge-Devices
We introduce Fact-Hash, a novel parameter-encoding method for training on-device neural radiance fields. Neural Radiance Fields (NeRF) have proven pivotal in 3D representations, but their applications are limited due to large computational resources. On-device training can open large application fields, providing strength in communication limitations, privacy concerns, and fast adaptation to a frequently changing scene. However, challenges such as limited resources (GPU memory, storage, and power) impede their deployment. To handle this, we introduce Fact-Hash, a novel parameter-encoding merging Tensor Factorization and Hash-encoding techniques. This integration offers two benefits: the use of rich high-resolution features and the few-shot robustness. In Fact-Hash, we project 3D coordinates into multiple lower-dimensional forms (2D or 1D) before applying the hash function and then aggregate them into a single feature. Comparative evaluations against state-of-the-art methods demonstrate Fact-Hash's superior memory efficiency, preserving quality and rendering speed. Fact-Hash saves memory usage by over one-third while maintaining the PSNR values compared to previous encoding methods. The on-device experiment validates the superiority of Fact-Hash compared to alternative positional encoding methods in computational efficiency and energy consumption. These findings highlight Fact-Hash as a promising solution to improve feature grid representation, address memory constraints, and improve quality in various applications.
- Open Access
HDR-Plenoxels: Self-Calibrating High Dynamic Range Radiance Fields
We propose high dynamic range (HDR) radiance fields, HDR-Plenoxels, that learn a plenoptic function of 3D HDR radiance fields, geometry information, and varying camera settings inherent in 2D low dynamic range (LDR) images. 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. To deal with various cameras in real-world scenarios, we introduce a tone mapping module that models the digital in-camera imaging pipeline (ISP) and disentangles radiometric settings. Our tone mapping module allows us to render by controlling the radiometric settings of each novel view. Finally, we build a multi-view dataset with varying camera conditions, which fits our problem setting. Our experiments show that HDR-Plenoxels can express detail and high-quality HDR novel views from only LDR images with various cameras.
Open AccessHDR-Plenoxels: Self-Calibrating High Dynamic Range Radiance Fields
We propose high dynamic range (HDR) radiance fields, HDR-Plenoxels, that learn a plenoptic function of 3D HDR radiance fields, geometry information, and varying camera settings inherent in 2D low dynamic range (LDR) images. 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. To deal with various cameras in real-world scenarios, we introduce a tone mapping module that models the digital in-camera imaging pipeline (ISP) and disentangles radiometric settings. Our tone mapping module allows us to render by controlling the radiometric settings of each novel view. Finally, we build a multi-view dataset with varying camera conditions, which fits our problem setting. Our experiments show that HDR-Plenoxels can express detail and high-quality HDR novel views from only LDR images with various cameras.
- 11 November 2022
- Published by Springer Nature
- p. 384-401
HDR-Plenoxels: Self-Calibrating High Dynamic Range Radiance Fields
- 11 November 2022
- Published by Springer Nature
- p. 384-401
HDR-Plenoxels: Self-Calibrating High Dynamic Range Radiance Fields
- Open Access
- 30 November 2020
- Published by Korean Institute of Electromagnetic Engineering and Science in The Journal of Korean Institute of Electromagnetic Engineering and Science
- Vol. 31 (11) , 986-989
Object Detection Using Stereo Vision and Radar Sensor Fusion
한국전자파학회논문지 (The Journal of Korean Institute of Electromagnetic Engineering and Science)
- 30 November 2020
- Published by Korean Institute of Electromagnetic Engineering and Science in The Journal of Korean Institute of Electromagnetic Engineering and Science
- Vol. 31 (11) , 986-989
Open AccessObject Detection Using Stereo Vision and Radar Sensor Fusion
한국전자파학회논문지 (The Journal of Korean Institute of Electromagnetic Engineering and Science)
- Open Access
- 30 November 2020
- Published by Korean Institute of Electromagnetic Engineering and Science in The Journal of Korean Institute of Electromagnetic Engineering and Science
- Vol. 31 (11) , 990-993
77 GHz Radar Transceiver with Horn Antenna on FR-4 PCB
한국전자파학회논문지 (The Journal of Korean Institute of Electromagnetic Engineering and Science)
- 30 November 2020
- Published by Korean Institute of Electromagnetic Engineering and Science in The Journal of Korean Institute of Electromagnetic Engineering and Science
- Vol. 31 (11) , 990-993
Open Access77 GHz Radar Transceiver with Horn Antenna on FR-4 PCB
한국전자파학회논문지 (The Journal of Korean Institute of Electromagnetic Engineering and Science)
- Open Access
- 20 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 213854-213865
79 GHz Active Array FMCW Radar System on Low-Cost FR-4 Substrates
This work, for the first time to our best knowledge, presents a W-band multi-channel frequency-modulated continuous-wave (FMCW) radar system on low-cost FR-4 substrates. A center-fed patch array antenna on an FR-4 substrate can achieve a maximum gain of 10.8 dBi. It is enabled by aperture coupled feeding using an on-chip feeder implemented on a CMOS radar transmitter (Tx) and receiver (Rx). The proposed radar system consists of one Tx and four Rx channels placed in the back cavity of a microstrip array antenna like an active array antenna system. Additionally, Tx and Rx chipsets include a wideband frequency multiplier with high multiplication ratio of 63, making it easy to distribute reference FMCW waveform synthesized at a very low frequency about 1.25 GHz for 79 GHz output using direct digital synthesis (DDS). Extending the number of channels and implementing various waveforms can be easily accomplished with very good phase noise. The functionality of the proposed radar system on low-cost substrates is confirmed by distance and angle measurements.
- 20 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 213854-213865
Open Access79 GHz Active Array FMCW Radar System on Low-Cost FR-4 Substrates
This work, for the first time to our best knowledge, presents a W-band multi-channel frequency-modulated continuous-wave (FMCW) radar system on low-cost FR-4 substrates. A center-fed patch array antenna on an FR-4 substrate can achieve a maximum gain of 10.8 dBi. It is enabled by aperture coupled feeding using an on-chip feeder implemented on a CMOS radar transmitter (Tx) and receiver (Rx). The proposed radar system consists of one Tx and four Rx channels placed in the back cavity of a microstrip array antenna like an active array antenna system. Additionally, Tx and Rx chipsets include a wideband frequency multiplier with high multiplication ratio of 63, making it easy to distribute reference FMCW waveform synthesized at a very low frequency about 1.25 GHz for 79 GHz output using direct digital synthesis (DDS). Extending the number of channels and implementing various waveforms can be easily accomplished with very good phase noise. The functionality of the proposed radar system on low-cost substrates is confirmed by distance and angle measurements.