CV

Reliability Engineer (Computer Vision)

Apple Inc.

Introduced a foundation model–based few-shot learning method for generalizable visual anomaly detection, enabling accurate cosmetic defect detection and quantification across products such as Vision Pro, AirPods, and Apple Watch without labeling or training.

Conputer Vision Algorithm Intern

Xiaomi Technology

Developing a video generation algorithm for autonomous driving scenarios. Our algorithm combines the merits of both 3D Gaussian splatting and denoising diffusion models to produce high-fidelity and physically consistent synthetic data.

Research Assistant

Shenzhen Research Institute of Big Data, The Chinese University of Hong Kong, Shenzhen

Built 3D colon reconstruction software that takes RGB endoscopy videos as input and reconstructs 3D surfaces for medical analysis.

Master of Electrical and Computer Engineering

Rice University

Computer Vision

• Introduction to deep machine learning
• Neural methods for image synthesis
• 3D vision
• Fundamentals of robotic manipulation
• Computer systems architecture

Bachelor of Engineering Degree in Automation

Sichuan University

Automation

• Computer Fundamentals and C Programming
• Data Structures & Algorithms
• Pattern Recognition
• Artificial Intelligence
• Machine Learning

TranSplat: Lighting-Consistent Cross-Scene Object Transfer with 3D Gaussian Splatting

We present TranSplat, a 3D scene rendering algorithm that enables realistic cross-scene object transfer (from a source to a target scene) based on the Gaussian Splatting framework. Our approach addresses two critical challenges: (1) precise 3D object extraction from the source scene, and (2) faithful relighting of the transferred object in the target scene without explicit material property estimation.

ORB-SfMLearner: ORB-Guided Self-supervised Visual Odometry with Selective Online Adaptation

ICRA 2025

We propose an Oriented FAST and Rotated BRIEF (ORB)-guided visual odometry with selective online adaptation named ORB-SfMLearner. The integration of ORB augmentation has improved the accuracy and explainability of our model’s pose estimation, while the proposed Selective Online Adaptation further enhanced its generalizability.

ToDER: Towards Colonoscopy Depth Estimation and Reconstruction with Geometry Constraint Adaptation

We propose a novel 3D reconstruction method for endoscope videos.

Three-stage binarization of color document images based on discrete wavelet transform and generative adversarial networks

Accepted by Knowledge-Based Systems

We propose a three-stage method using generative adversarial networks (GANs) for the degraded color document images binarization that achieves state-of-the-art performance.

A self-attention-embedded deep learning model for phasor measurement unit-based post-fault transient stability prediction.

2022 Asian Conference on Frontiers of Power and Energy (ACFPE)

We propose a novel deep learning model embedded with self-attention layers for transient stability assessment. Our model is able to achieve accurate and also explainable results.

Scene Editing with 3D Gaussian Splatting

Deep Visual Odometry with Online Adaptation

Monocular 3D reconstruction and measurement Application

Stability analysis of AC / DC power grid based on Artificial Intelligence

Outstanding Graduate of Sichuan University

National Undergraduate Electronics Design Third prize

Comprehensive scholarship of Sichuan University Second prize