chusheng0505.github.io

Collections of My Projects

Follow me on GitHub

~~ Welcome to GitHub Pages ~~


Introductions:

Name : Chu Sheng,Tan
Educations :

  1. (Bachelor) Taiwan, National ChungCheng University (Department of Physics)
  2. (Master) Taiwan, National ChengChi University (Department of Stastitics)
  3. Taiwan, AI Academy 台灣人工智慧學校 九期學員

Current Jobs : AI & Software Engineer
Languages : Python
Fields :
1.Theorical Calculations of Mathematics
2.WebScrapping
3.Statistical Analaysis and Modeling
4.Computer Vision with Machine Learning and Deep Learning
5.NLP model
Framework : Tensorflow / Pytorch / PySpark / Dask
Working Experiences :
1.(Data Analyst and Engineering) Little King Ind. Co., Ltd. (2019/08 - 2022/04 , 2 years and 9 months)
2.(AI & Software Engineer) Quanta Computer lnc. (2022/12 - )
Email : noble1993chusheng@hotmail.com


Project 1 : Classifying-text-in-Amazon-Captcha-by-using-KNN

aims : Try to predict the captcha appeared during webscrapping.
model used : KNN (k-nearest neighbour classifications)
inputs : images of captcha
output : text of captcha
accuracy : 94 - 96 %

Project 2 : Image-Captions-and-Generating-Images

aims : Try to search those relatively images through descriptions.
model used : pseudo siamese network(pretrained Bert + CNN);facebook’s Hubert ; CP-GAN
inputs : sentences or speech(in .wav format)
output : images stastified conditions

Project 3 : Tesseract-OCR-and-Automated-Files-Processing

aims : To save key in time.
model used : Tesseract-OCR ( Google )
inputs : pdf files
output : .xlsx or .csv files which are extracted from images and sorted.

Project 4 : Forecast-Modeling-and-Predictions

aims : To predict the demand in future.
model used : Time Series Model ,eg : ARIMA / SARIMA / GARCH and etc
statistical test : Augmented Dickey-Fuller Test ; Durbin Watson Value ; Model AIC
inputs : numerical historical data
output : numerical predictions data

Project 5 : Object Detections and Text Recognitions

aims : To detect the locations of target(including texts and numbers) and recognize the text on it.
model used : YoloV3 ; CNN
inputs : images
output : results of text

Project 6 : [Paper Research]

Classifications :

  1. DarkNet-53
  2. SE-ResNet50

Segmentations :

  1. Bi_Direc_ConvLSTM
  2. TransUnet
  3. RefineNet
  4. MultiResUnet