About
My name is Wei-Hsiang Liao and you may call me David. I recently graduated from the Artificial Intelligence (AI) Degree Program at the National Yang Ming Chiao Tung University (NYCU). During my time in graduate school, my research focused on robotics, self-driving cars, mathematical algorithms, and deep learning. Additionally, I hold a bachelor's degree in electrical engineering, giving me a solid foundation in electrical knowledge.
I am passionate about coding and researching algorithms, and am proficient in programming languages such as C/C++ and Python. My master's thesis was accepted for publication at the ICRA (IEEE International Conference on Robotics and Automation), and I am eager to apply my knowledge and skills in a professional setting.
Besides work, I love to do exercises, such as weight training, running, and swimming. I also love to do outdoor activities with friends.
Skills
C/C++
Python
Shell
Others
Linux
Robot Operating System (ROS)
Git
Pytorch
Projects
Self Driving Car Localization Using ICP Algorithm
Localize self-driving cars using Iterative Closest Point (ICP)
Github
SIFT Image Stitching
We implement SIFT to find the correspondes between two image and stitch them.
Github
Publications
GNN-based Point Cloud Maps Feature Extraction and Residual Feature Fusion for 3D Object Detection
LiDAR detection of long-range vehicles is challenging because very few and sparse points are measured in long distances and vehicles with similar shapes of targets could lead to false positives easily. To tackle these challenges, we take the environment information (HD maps) into account and construct a GNN-based feature extraction of point cloud maps to increase the receptive fields of learning map features. Residual feature fusion is proposed to fuse the features from PVRCNN and the map features from GNN. Our approach is evaluated on NuScenes dataset. It achieves a 24.78% average precision improvement for long-range objects at 40-50 meters, the farthest areas with ground truth annotation. Our approach also has a 4.22% reduction of false positives in the entire sensing areas.
Reconstruction and Synthesis of Lidar Point Clouds of Spray
Lidars are commonly used on autonomous vehicles, but their performance can be significantly affected by adverse weather. Therefore, we produce a lot of spray data and combine them into lidar object detection model to erase the effect of spray. In this paper, we propose the first data-driven method combined with simulation to reconstruct and synthesize spray data. We compare the performance of vehicle detection models trained with and without augmented data. The model trained with augmented data achieves significant performance improvement given real-world spray-affected point cloud data.
