吳柏威(Po-Wei Wu)

  maya6282@gmail.com
  +886 987-345-097

Education

Master,National Taiwan University

2017 / 9 - 2019 / 6

Institute of computer science and information engineer

Research Fields: Deep learning, Generative Adversarial network

Bachelor,National Tsing Hua University

2013 / 9 - 2017 / 6

Computer science

Skills


Language

  • C/C++
  • Java
  • Python
  • Javascript
  • HTML / CSS


Machine/Deep Learning

  • Keras
  • Tensorflow
  • Pytorch
  • Scikit-learn
  • MultiProcessing


Image processing

  • OpenCV
  • Scikit-image
  • Scikit-video


Web Development

  • MYSQL / mongoDB
  • Flask / NodeJs
  • Jquery / React / Angular
  • Bootstrap / Materialize

Work Experience

Microsoft Taiwan,2019 / 3 ~ 2018 / 7

Intern : Research Assistant

  • Using GAN to remove glasses to improve face recognition system accuracy
  • Collection face recognition dataset and doing data cleaning

Dcard Tech,2018 / 7 ~ 2018 / 8

Intern : Data Scientist

  • Develop deep learning model to detect malicious message (92%) & erotic pictures (95%)
  • Develop style transfer system to make the special effect on stories on Dcard

Microsoft Taiwan,2016 / 7 ~ 2017 / 7

Intern : Service Team

  • Assist with writing RA document & SA document
  • Assist with developing API in ETC
  • Assist with developing frontend interface in King’s town bank
  • Develop Software of POC such as smart healthcare system

ITE Tech,2015 / 7 ~ 2015 / 8

Intern


Publication

RA-GAN: Multi-domain Image-to-Image Translation via Relative Attributes (ICCV 2019)

An Evaluation on Bitcoin Address Classification based on Transaction History Summarization (ICBC 2019)

Deep Reinforcement Learning for Playing 2.5D Fighting Games (ICIP 2018)

Projects

Relative-Attribute GAN 

Projects 01 00@2x 9a5a97084c8e268ae3725d8e03654d1815ae2eb597b59c292ae4d0beffa7b703

  • Use deep learning to give the face variable spatial effect.
  • More flexible, practical, scalable than previous work

GanExample

Projects 01 00@2x 9a5a97084c8e268ae3725d8e03654d1815ae2eb597b59c292ae4d0beffa7b703

Research classical Generative Adversarial Network model and reproduce it

lf2gymProjects 01 00@2x 9a5a97084c8e268ae3725d8e03654d1815ae2eb597b59c292ae4d0beffa7b703

  • Using novel reinforcement learning model defeat original rule-based AI
  • My job is modifying the original LF2 project to fit our training environment

Bitcoin Address Classification

Projects 01 00@2x 9a5a97084c8e268ae3725d8e03654d1815ae2eb597b59c292ae4d0beffa7b703

Using new feature and NN improve the classification accuracy

Other Projects

JOB-EXPO

It is a project used in Job expo held by Microsoft Taiwan

ADLxMLDS2017

The projects in deep learning course in NTU

I got A+ in this class

CudaLecturePractice

Research classical parallel algorithm and implement it with CUDA

celeba-hq-modified

Modified h5tool.py make user getting celeba-HQ easier

Prize

The 3rd Contest of NTHU/ITRI Open Source Incubation Program Merit Award

2016 / 3

Award unit : Networking and Multimedia Systems Lab, NTHU

Design a open source project using android p2p communication system.

Presidential Award

2015 / 7

Award unit : National Tsing Hua University

Awarded to top 5% students in class each semester. I got the first place of 124 students in CS department.

Powered by CakeResumePowered by CakeResume