Chengjie Huang

Title: Chengjie Huang: Inventor in 3D Scene Representation and Object Detection

Introduction

Chengjie Huang is an inventor based in Kitchener, Canada, and works at Huawei Technologies Canada Co., Ltd. His patent work focuses on 3D point-cloud processing, object detection, and methods for generating training data and representations of driving scenes. Huang is associated with 2 patents.

Latest Patents

One of Huang’s latest patents, “Method and a Server for Generating Training Data for Training an Object Detector,” describes a way to create training data for fine-tuning an object detector. The method selects a quasi-stationary bounding box from a sequence of labelled 3D point clouds, relocates an object’s bounding box to a corresponding position in another point cloud, and stores the relabelled data for training.

His other latest patent, “Methods and Systems for Generating a Three-Dimensional Representation of a Scene,” concerns building 3D scene representations from successive point clouds. It estimates an object’s position, uses the object’s dynamic characteristics to determine how many preceding point clouds to consider, and aggregates data points representing the object. The approach may support more accurate interpretation of driving scenarios.

Career Highlights

Huang’s listed inventions address practical challenges in developing object detectors and reconstructing scenes from 3D sensor data. Together, they reflect an emphasis on using sequences of point clouds to improve training data and track objects across changing scenes.

Collaborations

Huang’s listed coworkers include Vahdat Abdelzad and Sean Albert Sedwards. His work is associated with Huawei Technologies Canada Co., Ltd.

Conclusion

Chengjie Huang’s patent portfolio highlights methods for preparing object-detection training data and generating 3D representations from point clouds, with potential applications in understanding driving environments.