湯沂達

畢業於交通大學應用數學系,前台灣人工智慧學校的AI工程師。

擔任過台灣人工智慧小聚講者;18組專案之技術指導;IEEE機器學習競賽第二名、TBrain競賽第五名。 

擅長於數學建模、機器學習、深度學習、電腦視覺、聲音信號處理、風格轉換。

閒暇時間會閱讀一些AI相關論文、線上論壇、新聞。


近期研究項目 Style Transfer, Text Driven Image Manipulation/Genreation, Neural Rendering, Network Pruning

Machine Learning Engineer
[email protected] / +886 917-622-216 / Taipei Taiwan


Link of this Page : https://www.cakeresume.com/changethewhat

專業能力


  • Programming Language: Python
  • Deep Learning : ONNX, PyTorch, TorchElastic, Tensorflow, Tensorflow Lite, CoralTPU, Keras, tfmot
  • Dev/ML Ops : Git, Linux, Docker, Google Apps Script, Google Drive API, Colab, AWS SageMaker, Weight And Bias
  • Applied Skills : Machine Learning, Deep Learning, Image Processing, Computer Vision, Parallel Programming, Signal Procressing, Mathematic Modeling, Optimization, Model Compression, Neural Style Transfer, Audio Classification

工作經歷

AI Academy, AI 工程師, Jun 2019 ~ Feb 2022

教授學員機器學習、深度學習知識;以技術指導的角色,幫助學員完成AI專案;參與外部 AI 競賽,累積公司名聲。

開發入學、內部考試系統、因應疫情之遠端點名系統。

閱讀論文、函式庫,並將之應用在比賽[1, 2]/專案指導[3]/教學文件、讀書會、Meidum、Meetup分享[4-9]/微型專案[10-13]/論文延伸方法提出[14]。

領導學員完成AI專案。(18組專案) [15]

Reference:

[1] (2019, Certificate) The 2nd Place of IEEE Big Data FEMH Voice Data Challenge 2019 (https://drive.google.com/drive/folders/175LPHjGcpp9qo01KIUqFNkw1PC7a3aO6?usp=sharing)


[2] (2021, Certificate) TBrain-Tomofun 狗音辨識 AI 百萬挑戰賽第五名(https://tbrain.trendmicro.com.tw/Competitions/Details/15)

(https://drive.google.com/drive/u/0/folders/18uk17_LipGpNLEe5f7SljISy0Lupny7j)


[3] (2021, Slide) AI專案分享
(https://slides.com/sin_dar_soup/ai#/6/1)


[4] (2019, Slide) 看見聲音的未來
(https://www.facebook.com/groups/datasci.dinner/permalink/2415658465412338/


[5] (2020, Slide) Re-Overview Model Compression and Acceleration (https://docs.google.com/presentation/d/1RINEfysY5H-ibBq_Rs6fTnF338gJGQjewyewfwqyas4/edit#slide=id.p)


[6] (2020, Slide) Transferred/Compact Convolutional Filters (https://slides.com/sin_dar_soup/deck#/transferredcompact-convolutional-filters)


[7] (2021, Talk) 從風格轉換到用文字操弄影像
From Style Trasfer to Text-Driven Image Manipulation

(video, slide)


[8] (2022, Talk) 初探類神經渲染

Neural Rendering, From Begineer

(videoslide)


[9] Medium (https://changethewhat.medium.com/)

(2021) 類神經影像藝術風格轉換系列筆記-基礎

(2021) 資料 : {(影像,文字)} : OpenAI 2021 一月
(2021) 用CLIP達成文字操弄/生成影像


[10] (2021, Side Project) Download-Google-Drive-Files-From-URLs

code : https://github.com/mistake0316/Download-Google-Drive-Files-From-URLs


[11] (2021, Side Project) Remote Sign System

# Using Google Apps Script to generate/handle sign periodically 

# no link


[12] (2021, Side Project) CLIPStyleTransfer

# Do Style Transfer with Text prompt

code : https://github.com/mistake0316/CLIPStyleTransfer

replicate : https://replicate.com/mistake0316/style-transfer-clip

project page : not release yet


[13] (2022, Side Project) Probability Awared Pruning

# A novel channel-wise pruning heuristic based on AdaIN's scale and bias

wanb : https://wandb.ai/mistake0316/ElascitPruningStyleTransfer

project page : not release yet


[14] Idea of an extension to EagleEye 

# EagleEye is an elegant heuristic to prune a model

# I proposed an idea extension to the authors in following github issue

 https://github.com/anonymous47823493/EagleEye/issues/33


[15] 指導專案

1.聲音的瑕疵檢測與分析

2.RF人體特定動作識別

3.黃光CD線寬預測

4.判斷鏡頭生產品質

5.智慧水表之用水模式辨識

6.以人工智慧為戲劇配樂

7.廢水新砂濾功能辨識系統

8.L5A 2B3T樹脂效率判定與餘命預警

9.Robot cable tray monitor system

10.影片人物的真偽判別 Anti-DeepFake

11.動態物件偵測尋找目標之無人載具

12.智慧手機架

13.購物網站訪客瀏覽行為分析

14. 以人工智慧估算電池健康度SOH

15. Hand Gestures Recognition with Millimeter Wave Radar

16. CNN模型應用於邊緣計算的手寫辨識系統實作

17. AI判斷骨齡

18. AI-Painter