Jungho Lee

I am a Ph.D candidate at Yonsei University in Seoul, where I work on computer vision and machine learning.

My primary areas of research are 3D computer vision techniques including 3D Gaussian Splatting (3D-GS) and Neural Radiance Fields (NeRF), and video understanding, specifically focusing on action recognition.

I'm always open to collaborations or suggestions. Please feel free to contact me if you have any questions or suggestions. :)

Email  /  CV  /  Linkedin  /  Google Scholar  /  Github

We propose a hierarchically decomposed graph convolution with a novel hierarchically decomposed graph, which consider the sematic correlation between the joints and the edges of the human skeleton.

We propose a novel Spatio-Temporal Curve Network (STC-Net) for skeleton-based action recognition, which consists of spatial modules with an spatio-temporal curve (STC) module and graph convolution with dilated kernels (DK-GC)

We propose a guided slot attention network to reinforce spatial structural information and obtain better foreground–background separation.

We propose the multi-scale structural graph convolutional network (MSS-GCN) for skeleton-based action recognition, the common intersection graph convolution (CI-GC) leverages the overlapped neighbor information between neighboring vertices for a given pair of root vertices.

We propose continous motion-aware blur kernel on 3D gaussian splatting utilizing 3D rigid transformation and neural ordinary differential function to reconstruct accurate 3D scene from blurry images with real-time rendering speed.

We propose novel blur kernel for motion estimation based on neural ordinary differential function to construct the deblurred neural radiance fields.

We propose novel bi-directional token-masking autoencoder (BTMAE) for referring image segmentation (RIS) to effectively utilize contextual information between language and visual features.

We propose enhanced deblurred neural radiance fields from sparse view settings for more practical applications considering real-world scenarios.

We propose a novel motion-as-option network by treating motion cues as optional and an adaptive output selection algorithm to adopt optimal prediction result at test time.

This website's source code is borrowed from jonbarron's website.

Last updated July 2023.