homata123 Goto Github PK

Hi there 👋

📫 How to reach me:

Github: https://github.com/homata123

Email: [email protected]

LinkedIn: https://www.linkedin.com/in/m%E1%BA%A1nh-th%E1%BA%AFng-b6323b19b/

My orientation and work mainly focused on Machine Learning,AI and i use Python as main language in my projects. I always feel happy to have chance to learn new knowledge and technologies. I can work independently or colaborate with team. Research and implement methods,models are my strengths .I have little experience and knowledges in building FE,BE or software but i can improve myself by learning from people around.

🔭 I’m currently working as a mentor on ZeroX,a technology startup company

Read my posts on blog Viblo.Asia about technology topic here : https://viblo.asia/u/homata123

I have ever been an AI internship to earn exprience in some companies, join in research works . Here are some representative projects which i have made,mostly in ML/CV field. (see more in my Github )

1. HUSTGRADE - A Scan-and-Grade exam score local web app built with OpenCV

An APP that will help the teachers get the results of answer sheet of the students faster and automatically . Just browse scan image,then let it grades the exam and return the number of correct/wrong marked. The enhancements compare with other same-purpose-apps are it can detetect all cases of wrong-marked as mark more than 1 sheet or not mark any sheet, it also let the exam-manager manage the volume threshold of ink line,if a mark is too dull/unclear ,it will be graded as wrong.

Tool used:

Python 3.7

OpenCV

Streamlit

Pillow

2. Innovated "Summarizing-and -Correcting Text" method using mixing-similarity-measures and Bert summarize model

This is my self-research project based on latest reseachs of other scientists in NLP field for solving the summarize-text problem. It simply get an input original text ( i.e a science paper,a newspaper before public) and return a summary text which can be specified the number of output sentence. It take advantages from some efficient similarity measure as Cosine and Jaccard to find the most meaningful sentences in a short time ,in additional the Bert model is used to find the key words in text. Therefore it can achieve the task with highly proper resources and accuracy. It is still in development progress but the result seems very potential.

Tool used:

Python 3.7

BERT model

Streamlit