Hand-craft pokemon yellow web

This project is for NTUT object oriented lab course. The rule is, using object oriented and easy framework (not a game engine, only loop) to make a comprehensive game. Finally, we spent 80 hours and got the 1st score in course.

intro web (play and github): https://nawanae.github.io/PokemonWebV2/


Technology stack: JS, css, html, canvas

OCR improvement by debluring mobile phone shot doc

This is the final project for Deep Learning and NLP Research Methodologies course, NTHU. We propose our debluring model to handle OCR blurry input. Hence, we adopt the SmartDocQA, a real-world phone took text images dataset, as our deblur dataset. After our aligment of dataset and training, finally we get better OCR accuracy.
report: https://drive.google.com/file/d/1YN7MRtP-H9nG6fFzkbsAzOiep2lWzLnj/view?usp=sharing
Technology stack: Tesseract OCR, DMPHN, pytorch, CBAM, CNN, deblury

Vehicle Trajectories Prediction by Machine Learning

The graduated project of NTUT CS dept. We cooperated with M.S. student to enhance the SGAN trajectories prediction model. Eventually, we get better predictions by extracting new features and attention in model.
Report: https://drive.google.com/file/d/1DOYCHNcQMpwV43imlYQUIwkmYVo7kjZe/view
Technology stack: pytorch, SGAN, LSTM

Tea Factory Website

This project is for NTUT web crouse final. We combine MDL with Bootstrap as our front-den. Adopted Firebase as DB system, hence the website have login and notice functions.

Github: https://github.com/a105590016/TeaWebFinal-Project

Technology stack: MDL design, BS4, firebase, javascript, html, css

Discord Language Learning Bot

The discord bot contains several NLP functions help users learn language. This project is for NTHU NLP lab crouse final project, finally, we get 2nd place in final competition.
Functions:
1. prefix, suffix, root search in your input vocabulary
2. vocabulary using frequency search
3. sentence confidence, emotion and subjectivity check
4. scrabble
5. daily vocabulary, prefix, root and suffix
https://github.com/kevin879275/NLP_FinalProject
Technology stack: discord.js, nodejs, python, linggle, word2vec, gensim, pattern

Milk Volume Prediction

In this project, we use a XGBoost and Res-regression model to try to handle the milk volume prediction problem, predict milk volume by datasets include cattle information, which is raised by the TPCIIA.
Results:
1. Res-regression is better than average and the others traditional model, and get RMSE 5.8.
2. XGB regressor well fit this problem because of the data uncompleted and the small amount, finally get RMSE 5.47.
Technology stack: pytorch, pandas, Res-regrssion, XGBoost