Jingyu Guo

Past Research

I have worked on 3D representation learning, generative modeling, and computer vision methods for understanding sports broadcast videos.

HUGE-Bench high-level UAV trajectory examples

HUGE-Bench: A Benchmark for High-Level UAV Vision-Language-Action Tasks

Jingyu Guo, Ziye Chen, Ziwen Li, Zhengqing Gao, Jiaxin Huang, Hanlue Zhang, Fengming Huang, Yu Yao, Tongliang Liu, and Mingming Gong

European Conference on Computer Vision (ECCV), 2026

TL;DR: HUGE-Bench evaluates whether UAV agents can follow concise high-level commands and execute safe multi-stage trajectories in digital twin scenes, using process-oriented and collision-aware metrics.

Project Page / Paper / Code

Hyper3D qualitative reconstruction comparison

Hyper3D: Efficient 3D Representation via Hybrid Triplane and Octree Feature for Enhanced 3D Shape Variational Auto-Encoders

Jingyu Guo, Sensen Gao, Jia-Wang Bian, Wanhu Sun, Heliang Zheng, Rongfei Jia, and Mingming Gong

arXiv preprint arXiv:2503.10403, 2025

TL;DR: Hyper3D combines octree-based mesh features with a hybrid triplane-grid latent space, improving 3D shape VAE reconstruction fidelity while keeping latent representations efficient.

Paper

Recognizing a Sequence of Events from Tennis Video Clips: Addressing Timestep Identification and Subtle Class Differences

Zhaoyu Liu, Jingyu Guo, Mo Wang, Ruicong Wang, Kan Jiang, and Jin Song Dong

2023 IEEE 28th Pacific Rim International Symposium on Dependable Computing (PRDC)

TL;DR: We propose an end-to-end network for temporally precise tennis event detection, addressing timestep identification and subtle class differences in broadcast clips.

Paper

Past Research

HUGE-Bench: A Benchmark for High-Level UAV Vision-Language-Action Tasks

Hyper3D: Efficient 3D Representation via Hybrid Triplane and Octree Feature for Enhanced 3D Shape Variational Auto-Encoders

Recognizing a Sequence of Events from Tennis Video Clips: Addressing Timestep Identification and Subtle Class Differences

Work Experience

Photography