Projects

PICO MR cloud-based mapping and localization — Visual Positioning Service (VPS).

High-performance Bundle Adjustment for large-scale scenes — Ceres CUDA, FastBA, MegBA.

Real-time sharing and localization of anchor maps across multiple devices, enabling multi-player MR games.

Vision-based, low-cost mapping and localization powering MR Spatial Anchors on PICO headsets.

Line-feature based VIO designed to improve re-localization performance on MR headsets.

Fuse cloud-based high-precision localization results into the on-device 6DoF tracking loop.

Use IMU data to compensate for the rolling-shutter "jelly" effect in RGB cameras under fast motion.

Multi-camera and lidar extrinsic calibration with a re-engineered workflow.

End-to-end OCR toolbox covering detection, recognition, post-processing, and TensorRT acceleration.

Camera-based semantic mapping for parking environments using line and marker features.

AGV-level multi-sensor fusion module for static and dynamic obstacle state estimation.

Evaluation toolboxes covering classification, keypoint detection, keypoint regression, and image-based 3D-box tasks.

Vehicle and cyclist 3D bounding box detection for the auto-driving perception stack.

Camera IPM-based parking space detection that emits 3D coordinates for downstream parking planning.

Implemented visual-inertial odometry for quadrotor pose and velocity estimation, with ROS simulation and real-flight tests.

Implemented polynomial trajectory generation for quadrotors and camera calibration using reprojection-error optimization.

Built a network for 3D-MNIST classification, 3D bounding box estimation, and depth estimation.

Implemented autoencoders, VAEs, and GAN variants, including image-to-image translation models.

Implemented Faster R-CNN with TensorFlow for object detection and classification using CIFAR datasets.

Implemented Structure from Motion, visual odometry, and bundle adjustment for quadrotor mapping.

Implemented object detection with GMMs, gesture recognition with HMMs, and reinforcement learning methods.

Implemented particle-filter SLAM using IMU prediction and laser-scan updates.

Implemented an Unscented Kalman Filter for orientation estimation using IMU data.

Implemented a Java API for fraction arithmetic.

Built face detection and replacement using color-based segmentation, PCA, and Face++.