Selected Projects

Comparison on fusion-based single image HDR enhancement

Assigned to ECE 613 Project.

Source Code

Odom-Laser Calibration Tool

Copyright Megvii Technology Ltd.


Here we implemented a easy-use tool for following parameters’ calibration: yaw-rotation and xy-translation between odom and 2D-LiDAR, and radius of odom’s wheels and axle between them. For more details and usage please reference to our project’s page.

For Chinese version introduction please reference to homepage of my colleague Dr. Yijia He.

Source Code

Continuous Localization and Mapping in low-light GPS-denied Environments

We propose a system for simultaneous localization and mapping in low-light environments based on multi-sensors fusion. Thermal Visual Odometry in our approach is obtained from a thermal camera’s multi-plane Homography decomposition with IMU Pose and laser altitude, and optimized by a Kalman Filter. In addition, based on multi-layers Grid Map, our system can create a robot-centric map fused by multi-plane extraction and laser scanner measurement.


plane detection 1

plane detection 2

grid map

landing point detection

Zhijie Wang, Yingcai Bi, Ben M. Chen, Continuous Localization and Mapping for Autonomous Navigation in low-light environments based on Multi-Sensors Fusion, IEEE International Conference on Control and Automation.

Coming Soon

Underwater Image Enhancer

We propose a novel architecture that can jointly extract local features and global features and elegantly fuse them together to perform underwater image enhancement based on the framework of conditional generative adversarial networks (cGANs) with residual blocks. The network is trained in an end-to-end fashion. Our approach is validated with subjective and objective tests and comparison against the state of the art, where we show great improvements.

Xiaodong Liu, Zhi Gao, Zhijie Wang, Yu Herng Tan, Yingcai Bi, Ben M. Chen, Underwater Image Enhancer based on Conditional GANs, IEEE Transactions on Circuits and Systems for Video Technology..

Coming Soon