wang shih hao

Computer graphics / deep learning RD • 台北市，TW • [email protected]

Currently working at htc, mainly responsible for computer graphics and mechanical learning related software development

Education

Education1，2008 年 7 月 - 2012 年 7 月

B.S. degree, National Sun Yat-sen University ,Majoring in Applied Mathematics

Educations 00 00@2x 7b148e1f93abab24dca94b4fdb4a0272c264e306a59e75c713416ce8b6bdd271

Education2，2012 年 7 月 - 2014 年 9 月

M.S. degree, National Chengchi University,Major in computer graphics

Educations 00 01@2x 4e43db9b699f13e23c72a9c02425174ede57b0cf1c591e8a1736b04cb30762e1

Experience

Working Experience1，2014/10 - 2017/8

3D model search site

Experiences 00 00@2x 504900dc09d82f711fdb54cf5763251cab5193a844856da978b2b8f3907ecd5a

Working Experience2，2017/8 - 2022/10

Mainly responsible for the development of MR and the application of Avatar

Experiences 00 01@2x 6daa47d1df70d315f129dcecdd562003831581c4cdd8f9c819aa22d4a24ae150

Working Experience3，2022/10 - now

Mainly responsible for the development of Avatar and mocap

M.S. degree:

Real-time Multiple Refraction by Using Image-Space Technology

English version:

My master's thesis is to implement real-time multi-layer refraction with shader.

The development process is detailed in link.

programming language: C++ ,glsl,opengl

yobi3d:

English version:

yobi3d is a 3D search engine that has been closed in 2019. For its introduction, you can refer to the YouTube video (https://www.youtube.com/watch?v=rU923b454l0). My main work in it has two parts below

1. Rendering of 3D models: Many 3D models captured by crawlers, in various formats (obj, str, fbx, ma….), I integrated them through the blender system and converted them into obj format. Use javascript and three.js to render to the web page as shown on the right.

2. Classification of 3D models: I use the D2 algorithm to vectorize the models, and after VOXELizing the models, I get the feature vector of each model, and classify the 3D models through machine learning, so I can find the model by model or use tag to find similar models.

HTC Proj1:

English version:

SRWork – The image captured through the front camera is used to implement MR

I use shaders to apply special effects, such as DEMO (https://www.youtube.com/watch?v=q8R7-hRoO0c), with sharpening, sketching, night vision goggles, thermal effects, etc.

Development Tools:Unity: Programming language: C#,CG shading language Development process: use VIVE SRWORK to get the image of see Throuht, and use CG shading language to process the image with special effects

HTC Proj2:

English version:

SRanipal – lip tracking ,eye tracking ,tongue tracking ViveProEye,

The lip camera module can track the facial expression system. I collected a lot of people's expressions, trained by machine learning, and finally presented them with the blenderShape technology in the 3D model, such as DEMO (https://www.youtube.com/watch?v=3ayhrtU2jwo&t=410s&ab_channel=CasandCharyVR).

Development tools: Unity, Deeplearning caffe, vs2015

Programming language: Python, C#, C++ Development process:

use Unity to design image capture tools, use Python and caffe for Deeplearning training, and finally use UNITY to implement a conten