Brian Wu
Computer Vision Developer

Brian Wu

Former senior engineer at NCS Singapore.
Work with Brian Work with a Computer Vision Developer

About

Brian Wu is an R&D engineer with 11+ years of experience designing and deploying computer vision and deep learning algorithms on embedded platforms such as FPGA and ARM. He specializes in FPGA/ZYNQ SoC design using Verilog HDL and GPU acceleration with CUDA and OpenGL. His expertise spans SLAM/Visual SLAM, ROS, VR/AR, stereo vision, 3D sensing, object detection, and motor control. He is proficient in C/C++, Python, JavaScript, Golang, and Verilog, and experienced with TensorFlow and Darknet. He also builds full-stack systems using Django, React/Vue/Next.js, MySQL, Node.js, and Go.

Employment

4 roles
Mar 2020 - Present 6 years 2 months
founder
DeCensorMedia
  • develop open source and decentralized personal social networking website which is free of censorship and data collection. it is more than just a website + chat app, it is a home server + cloud server based IOT system that enables one can send message or video call to people in any corner of the world,and promote his business with this website as well. Since it is an home IOT system, one can manage home devices remotely,and also off load a lot of time consuming programs from cellphone to home server.
  • website: https://decensormedia.org/
  • intro video in English: https://youtu.be/4yhPMDgDkT4
  • intro video in Chinese: https://youtu.be/tz41zJIZ_U4
  • intro ppt : https://decensormedia.org/ppt


Jun 2018 - Dec 2019 1 year 6 months
senior software engineer
NCS Singapore
  • develop fall prevention algorithm and prototype with 3D sensor. In this project I detect the patient’s 3D sensor using openpose package and 3D sensor. Openpose generate a 2D skeleton of a person, with 3D sensor it can extend to 3D skeleton, base on the pose and position of the patient on the bed we can tell its current position and predict his/her next position and pose, and trigger alarm if we predict he is trying to get off the bed.
  • train face detection and optimize for arm platform.This project is to calculate the average time of a visitor spending from entering the entrance to leaving the exit, so that the administration can decide how many counter should open to best balance the manpower and the customer’s waiting time in the queue. Face detection is carried out in edge device, recognition in the back-end server.


May 2017 - Jun 2018 1 year 1 month
founder
3Deepercept
  • In this project I develop C++ version of algorithm to verify the ideas on computer and develop a first version prototype, then deploy the C++ algorithm to verilog HDL and run on FPGA(Xilinx zynq) which is the second version prototype. I design the linux device driver for zynq to bring up the HDMI display so the disparity map can be seen on screen, transmit the disparity map to host computer via usb as well. For object detection I use YOLO!
  • I have opensourced this project in github: https://github.com/brianwchh/grassrootsstartup-ComputerVsion-zynq


Dec 2015 - Oct 2016 10 months
senior embedded software engineer
go6D shenzhen
  • in this project I design the algorithms to stitch two images from two fisheye camera to generate one panoramic image. In the prototype stage I use openGL to do the stitching and rendering for perspective viewing a desired portion of the image. In later version stitching is implemented in CUDA.
  • video demo :  https://youtu.be/HCIf8AKpDgk


Education

Sep 2003 - Jul 2007 3 years 10 months
bechelor, Physics
yichun university

Portfolio

1 item